Benchmark Confidence & Contamination Flags

8 of 218 ranked models currently meet the high-confidence threshold. The confidence indicator (1-4 dots) shows how much sourced benchmark coverage supports each score, while the table separates sourced, provisional, and generated evidence.

●●●●High

7+ categories, 20+ non-generated benchmarks

●●●○Good

5+ categories, 12+ non-generated benchmarks

●●○○Moderate

3+ categories, 8+ non-generated benchmarks

●○○○Low / Estimated

Limited sourced data, score is estimated

Confidence Distribution (Ranked Models)

High (4%)

Good (11%)

Moderate (13%)

157

Low / Estimated (72%)

How BenchLM Scores Work

Verified, provisional, and generated

Each benchmark value is tagged as manual (a hand-entered public row) or generated (inferred from related models). Generated rows are excluded from all public ranking logic. Manual rows are now split again into sourced rows for the verified leaderboard and source-unverified rows that can still appear in provisional mode.

Ranking Eligibility

A model must have at least 8 qualifying benchmarks across 2+ categories to rank in a lane. The provisional leaderboard uses rankable non-generated rows; the verified leaderboard uses sourced rows only. Models below the threshold are shown as tracked but unranked.

Category Eligibility

For category leaderboards, a model needs qualifying scores on at least half of the weighted benchmarks in that category. BenchLM computes this separately for provisional and verified ranking so sparse exact-source coverage cannot silently borrow strength from provisional rows.

Display-Only Benchmarks

Some benchmarks (MMLU, BBH, HumanEval, older AIME/HMMT variants) are shown for context but don't affect scoring. These are either saturated (top models all score 97%+) or have been superseded by harder versions.

218 models match these filters.

Model	Confidence	Prov. score	Sourced	Rankable	Generated	Coverage
Qwen3.7 Plus Alibaba	●●●●High	66.02	50	50	0	100%
Claude Opus 4.5 Anthropic	●●●●High	63.63	41	41	0	100%
Kimi K2.5 Moonshot AI	●●●●High	58.89	40	40	0	100%
Qwen3.6 Plus Alibaba	●●●●High	64.79	39	39	0	100%
Qwen3.5 397B Alibaba	●●●●High	57.46	35	35	0	100%
GLM-5 Z.AI	●●●●High	65.61	34	34	0	100%
Qwen3.7 Max Alibaba	●●●●High	71.6	32	32	0	100%
Gemini 3.5 Flash Google	●●●●High	64.67	23	23	0	100%
Claude Opus 5 Anthropic	●●●○Good	83.07	59	59	0	100%
Qwen3.6-35B-A3B Alibaba	●●●○Good	51.42	39	39	0	100%
Qwen3.6-27B Alibaba	●●●○Good	53.73	36	36	0	100%
GPT-5.4 OpenAI	●●●○Good	73.36	32	32	0	100%
Claude Opus 4.8 Anthropic	●●●○Good	76.58	31	31	0	100%
Claude Opus 4.6 Anthropic	●●●○Good	67.98	30	30	0	100%
GPT-5.5 OpenAI	●●●○Good	73.37	29	29	0	100%
Kimi K2.6 Moonshot AI	●●●○Good	60.21	29	29	0	100%
GPT-5.6 Sol OpenAI	●●●○Good	82	23	23	0	100%
Muse Spark Meta	●●●○Good	70.81	22	22	0	100%
GPT-5.6 Terra OpenAI	●●●○Good	72.91	21	21	0	100%
GPT-5.6 Luna OpenAI	●●●○Good	67.29	21	21	0	100%
Gemini 3.1 Pro Google	●●●○Good	55.98	20	20	0	100%
Claude Opus 4.7 (Adaptive) Anthropic	●●●○Good	72.13	20	20	0	100%
Claude Sonnet 4.6 Anthropic	●●●○Good	64.53	18	18	0	100%
Inkling-Small Thinking Machines Lab	●●●○Good	65.54	18	18	0	100%
Nemotron 3 Ultra NVIDIA	●●●○Good	46.67	17	17	0	100%
Inkling Thinking Machines Lab	●●●○Good	67.04	16	16	0	100%
Ling 3.0 Flash InclusionAI	●●●○Good	53.78	16	16	0	100%
GPT-5.2 OpenAI	●●●○Good	58.24	14	14	0	100%
Qwen3.5-122B-A10B Alibaba	●●●○Good	59.64	13	13	0	100%
Qwen3.5-27B Alibaba	●●●○Good	59.82	13	13	0	100%
MAI-Thinking-1 Microsoft	●●●○Good	50.93	13	13	0	100%
Qwen3.5-35B-A3B Alibaba	●●●○Good	56.15	13	13	0	100%
GPT-5.4 mini OpenAI	●●●○Good	56.86	12	12	0	100%
Qwen3.8 Max Alibaba	●●○○Moderate	79.91	40	9	0	444%
Kimi K3 Moonshot AI	●●○○Moderate	80.5	29	9	0	322%
DeepSeek V4 Pro 0813 DeepSeek	●●○○Moderate	61.16	23	23	0	100%
Muse Spark 1.1 Meta	●●○○Moderate	76.88	19	19	0	100%
Grok 4.20 xAI	●●○○Moderate	55.19	17	17	0	100%
Claude Sonnet 5 Anthropic	●●○○Moderate	64.78	16	16	0	100%
GLM-5.1 Z.AI	●●○○Moderate	66.91	16	16	0	100%
GLM-5.2 Z.AI	●●○○Moderate	63.34	15	15	0	100%
MiniMax M3 MiniMax	●●○○Moderate	68.74	15	15	0	100%
Claude Mythos 5 Anthropic	●●○○Moderate	83.21	14	14	0	100%
Nemotron 3 Nano Omni 30B A3B NVIDIA	●●○○Moderate	44.4	14	14	0	100%
Muse Glimmer 30B Meta	●●○○Moderate	52.5	13	8	0	163%
MiniCPM5-1B OpenBMB	●●○○Moderate	11.36	13	13	0	100%
Nemotron 3.5 Lightning 30B A3B NVFP4 NVIDIA	●●○○Moderate	26.79	13	13	0	100%
Claude Fable 5 Anthropic	●●○○Moderate	82.96	12	12	0	100%
Gemini 3 Pro Google	●●○○Moderate	67.37	11	11	0	100%
GPT-5.4 nano OpenAI	●●○○Moderate	66.59	11	11	0	100%
GPT-5.4 Pro OpenAI	●●○○Moderate	61.26	11	11	0	100%
Gemma 4 12B Google	●●○○Moderate	47.32	11	11	0	100%
ZAYA1-8B Zyphra	●●○○Moderate	30.96	10	10	0	100%
GLM-4.7 Z.AI	●●○○Moderate	60.66	9	9	0	100%
Grok 4.5 xAI	●●○○Moderate	75.42	9	9	0	100%
Qwen 3.6 Max (preview) Alibaba	●●○○Moderate	59.99	9	9	0	100%
Claude Sonnet 4.5 Anthropic	●●○○Moderate	54.16	9	9	0	100%
Gemini 3.7 Flash Google	●●○○Moderate	61.36	8	8	0	100%
Kimi K2.5 (Reasoning) Moonshot AI	●●○○Moderate	59.81	8	8	0	100%
MiMo-V2.5 Xiaomi	●●○○Moderate	59.16	8	8	0	100%
Hy3 Preview Tencent	●●○○Moderate	43.96	8	8	0	100%
MiniMax M2.7 MiniMax	●○○○Low / Estimated	~63.06	16	16	0	100%
Step 3.7 Flash StepFun	●○○○Low / Estimated	~50.91	10	10	0	100%
GPT-5.3 Codex OpenAI	●○○○Low / Estimated	~65.93	7	7	0	100%
MiMo-V2.5-Pro Xiaomi	●○○○Low / Estimated	~69.21	7	7	0	100%
GPT-5.5 Pro OpenAI	●○○○Low / Estimated	~64.17	6	6	0	100%
Grok 4.6 xAI	●○○○Low / Estimated	~63.41	6	0	0	0%
Agents-A1 InternScience	●○○○Low / Estimated	~60.79	6	6	0	100%
Claude Opus 4.7 Anthropic	●○○○Low / Estimated	~71.87	6	6	0	100%
DeepSeek V3.2 DeepSeek	●○○○Low / Estimated	~54.67	6	6	0	100%
Grok 4.3 xAI	●○○○Low / Estimated	~63.95	6	6	0	100%
GPT-4.1 OpenAI	●○○○Low / Estimated	~50.84	6	6	0	100%
Gemma 4 31B Google	●○○○Low / Estimated	~60.15	6	6	0	100%
LFM2.5-8B-A1B LiquidAI	●○○○Low / Estimated	~41.62	6	6	0	100%
Gemini 3.5 Flash-Lite Google	●○○○Low / Estimated	~65.29	5	5	0	100%
Gemini 3 Flash Google	●○○○Low / Estimated	~59.74	5	5	0	100%
Claude Haiku 4.5 Anthropic	●○○○Low / Estimated	~57.06	5	5	0	100%
o3-mini OpenAI	●○○○Low / Estimated	~46.68	5	5	0	100%
GPT-4.1 mini OpenAI	●○○○Low / Estimated	~44.14	5	5	0	100%
Gemini 2.5 Pro Google	●○○○Low / Estimated	~56.63	5	5	0	100%
o1 OpenAI	●○○○Low / Estimated	~48.17	4	4	0	100%
Qwen3 235B 2507 Alibaba	●○○○Low / Estimated	~56.45	4	4	0	100%
Qwen3.5 Plus Alibaba	●○○○Low / Estimated	~47.47	4	4	0	100%
Gemma 4 26B A4B Google	●○○○Low / Estimated	~57.16	4	4	0	100%
GPT-4.1 nano OpenAI	●○○○Low / Estimated	~42.2	4	4	0	100%
Trinity-Large-Preview Arcee AI	●○○○Low / Estimated	~56.39	4	4	0	100%
Trinity-Large-Thinking Arcee AI	●○○○Low / Estimated	~47.71	4	4	0	100%
GPT-5.1 OpenAI	●○○○Low / Estimated	~53.54	3	3	0	100%
Grok 4 xAI	●○○○Low / Estimated	~59.61	3	3	0	100%
GLM-4.6 Z.AI	●○○○Low / Estimated	~54.34	3	3	0	100%
Ling 2.6 Flash InclusionAI	●○○○Low / Estimated	~44.08	3	3	0	100%
Command A+ Cohere	●○○○Low / Estimated	~47.56	3	3	0	100%
Gemini 3.6 Flash Google	●○○○Low / Estimated	~75.54	2	2	0	100%
Exaone 4.0 32B LG AI Research	●○○○Low / Estimated	~40.65	2	2	0	100%
Claude 4.1 Opus Anthropic	●○○○Low / Estimated	~44.99	2	2	0	100%
Claude 4 Sonnet Anthropic	●○○○Low / Estimated	~41.97	2	2	0	100%
Gemini 3.1 Flash-Lite Google	●○○○Low / Estimated	~50.51	2	2	0	100%
o4-mini (high) OpenAI	●○○○Low / Estimated	~50.37	2	2	0	100%
Kimi K2 Moonshot AI	●○○○Low / Estimated	~26.32	2	2	0	100%
o3 OpenAI	●○○○Low / Estimated	~47.01	2	2	0	100%
Qwen3 235B 2507 (Reasoning) Alibaba	●○○○Low / Estimated	~58.47	2	2	0	100%
Gemini 2.5 Flash Google	●○○○Low / Estimated	~47.45	2	2	0	100%
Qwen3.5 Flash Alibaba	●○○○Low / Estimated	~47.27	2	2	0	100%
Grok 3 [Beta] xAI	●○○○Low / Estimated	~40.04	2	2	0	100%
Claude 3.5 Sonnet Anthropic	●○○○Low / Estimated	~48.11	2	2	0	100%
GPT-5 (high) OpenAI	●○○○Low / Estimated	~59.08	2	2	0	100%
GPT-5.2-Codex OpenAI	●○○○Low / Estimated	~58.19	2	2	0	100%
Kimi K2.7 Code Moonshot AI	●○○○Low / Estimated	~54.34	2	2	0	100%
GPT-5.1-Codex OpenAI	●○○○Low / Estimated	~52.77	2	2	0	100%
MiMo-V2-Flash Xiaomi	●○○○Low / Estimated	~53.3	1	1	0	100%
Gemini 3 Pro Deep Think Google	●○○○Low / Estimated	~61.79	1	1	0	100%
DeepSeek V3 DeepSeek	●○○○Low / Estimated	~44.28	1	1	0	100%
GPT-4o OpenAI	●○○○Low / Estimated	~40.81	1	1	0	100%
Llama 4 Scout Meta	●○○○Low / Estimated	~39.39	1	1	0	100%
Llama 4 Maverick Meta	●○○○Low / Estimated	~22.76	1	1	0	100%
Claude Opus 4.6 (Adaptive) Anthropic	●○○○Low / Estimated	~64.72	1	1	0	100%
GLM-5 (Reasoning) Z.AI	●○○○Low / Estimated	~60.25	1	1	0	100%
GLM-5-Turbo Z.AI	●○○○Low / Estimated	~66.08	1	1	0	100%
MiMo-V2-Pro Xiaomi	●○○○Low / Estimated	~66.94	1	1	0	100%
GLM-5V-Turbo Z.AI	●○○○Low / Estimated	~62.28	1	1	0	100%
Grok 4.1 Fast (Reasoning) xAI	●○○○Low / Estimated	~59.71	1	1	0	100%
MiMo-V2-Omni Xiaomi	●○○○Low / Estimated	~62.36	1	1	0	100%
DeepSeek V3.2 (Thinking) DeepSeek	●○○○Low / Estimated	~58.59	1	1	0	100%
Grok 4 Fast (Reasoning) xAI	●○○○Low / Estimated	~55.8	1	1	0	100%
MiniMax M2.5 MiniMax	●○○○Low / Estimated	~58.68	1	1	48	2%
GPT-OSS 120B OpenAI	●○○○Low / Estimated	~49.23	1	1	0	100%
GPT-5.1-Codex-Max OpenAI	●○○○Low / Estimated	~54.9	1	1	0	100%
GPT-OSS 20B OpenAI	●○○○Low / Estimated	~42.2	1	1	0	100%
Nemotron 3 Super 100B NVIDIA	●○○○Low / Estimated	~50.47	1	1	0	100%
GPT-5 mini OpenAI	●○○○Low / Estimated	~43.08	1	1	60	2%
Qwen3 Max Alibaba	●○○○Low / Estimated	~48.26	1	1	0	100%
Claude Opus 4.5 Thinking Anthropic	●○○○Low / Estimated	~57.46	1	1	0	100%
Muse Spark 1.2 Meta	●○○○Low / Estimated	~61.67	0	0	0	0%
Grok 4.1 xAI	●○○○Low / Estimated	~60.43	0	0	0	0%
Hy3 Tencent	●○○○Low / Estimated	~68.02	0	0	0	0%
Qwen3.5 397B (Reasoning) Alibaba	●○○○Low / Estimated	~59.96	0	0	0	0%
GPT-5.2 Instant OpenAI	●○○○Low / Estimated	~59.47	0	0	32	0%
GPT-5.2 Pro OpenAI	●○○○Low / Estimated	~66.87	0	0	32	0%
GPT-5.3 Instant OpenAI	●○○○Low / Estimated	~59.36	0	0	35	0%
Grok 4.1 Fast xAI	●○○○Low / Estimated	~50.57	0	0	0	0%
DeepSeek V3.1 (Reasoning) DeepSeek	●○○○Low / Estimated	~52.72	0	0	0	0%
DeepSeek V3.1 DeepSeek	●○○○Low / Estimated	~52.9	0	0	0	0%
Mistral Large 3 Mistral	●○○○Low / Estimated	~49.54	0	0	0	0%
GLM-4.5 Z.AI	●○○○Low / Estimated	~58.02	0	0	0	0%
DeepSeek-R1 DeepSeek	●○○○Low / Estimated	~50.95	0	0	0	0%
GPT-5.3-Codex-Spark OpenAI	●○○○Low / Estimated	~57.36	0	0	44	0%
Step 3.5 Flash StepFun	●○○○Low / Estimated	~54.29	0	0	50	0%
o1-preview OpenAI	●○○○Low / Estimated	~48.42	0	0	0	0%
GLM-4.5-Air Z.AI	●○○○Low / Estimated	~46.99	0	0	0	0%
GLM-4.7-Flash Z.AI	●○○○Low / Estimated	~50.46	0	0	45	0%
Gemma 3 27B Google	●○○○Low / Estimated	~41.04	0	0	0	0%
Mistral Small 4 Mistral	●○○○Low / Estimated	~46.31	0	0	0	0%
DeepSeek LLM 2.0 DeepSeek	●○○○Low / Estimated	~54.95	0	0	0	0%
Mercury 2 Inception	●○○○Low / Estimated	~48.14	0	0	55	0%
GPT-5 (medium) OpenAI	●○○○Low / Estimated	~54.29	0	0	0	0%
Claude 3 Opus Anthropic	●○○○Low / Estimated	~40.43	0	0	0	0%
Nemotron 3 Nano 30B NVIDIA	●○○○Low / Estimated	~53.36	0	0	0	0%
GPT-4o mini OpenAI	●○○○Low / Estimated	~37.22	0	0	0	0%
Mistral Large 2 Mistral	●○○○Low / Estimated	~41.96	0	0	0	0%
Qwen2.5-72B Alibaba	●○○○Low / Estimated	~52.56	0	0	0	0%
Llama 3.1 405B Meta	●○○○Low / Estimated	~52.11	0	0	0	0%
Nemotron 3 Super 120B A12B NVIDIA	●○○○Low / Estimated	~51.37	0	0	49	0%
Llama 3 70B Meta	●○○○Low / Estimated	~51.22	0	0	0	0%
Qwen2.5 Coder 32B Instruct Alibaba	●○○○Low / Estimated	~34.03	0	0	0	0%
DeepSeek Coder 2.0 DeepSeek	●○○○Low / Estimated	~50.67	0	0	0	0%
Seed 1.6 ByteDance	●○○○Low / Estimated	~50.62	0	0	32	0%
Gemini 1.5 Pro Google	●○○○Low / Estimated	~35	0	0	0	0%
Phi-4 Microsoft	●○○○Low / Estimated	~22.41	0	0	0	0%
Qwen2.5-1M Alibaba	●○○○Low / Estimated	~50.27	0	0	0	0%
DeepSeekMath V2 DeepSeek	●○○○Low / Estimated	~50.27	0	0	0	0%
Seed-2.0-Lite ByteDance	●○○○Low / Estimated	~50.22	0	0	32	0%
Ministral 3 14B (Reasoning) Mistral	●○○○Low / Estimated	~49.73	0	0	48	0%
o3-pro OpenAI	●○○○Low / Estimated	~47.43	0	0	0	0%
Ministral 3 14B Mistral	●○○○Low / Estimated	~33.89	0	0	48	0%
Aion-2.0 Aion Labs	●○○○Low / Estimated	~49.04	0	0	32	0%
Mixtral 8x22B Instruct v0.1 Mistral	●○○○Low / Estimated	~48.94	0	0	0	0%
Grok Code Fast 1 xAI	●○○○Low / Estimated	~37.79	0	0	0	0%
GPT-4 Turbo OpenAI	●○○○Low / Estimated	~26.71	0	0	0	0%
Z-1 Z	●○○○Low / Estimated	~45.49	0	0	0	0%
Seed 1.6 Flash ByteDance	●○○○Low / Estimated	~45.44	0	0	32	0%
Nemotron-4 15B NVIDIA	●○○○Low / Estimated	~45.44	0	0	0	0%
Mistral 8x7B Mistral	●○○○Low / Estimated	~45.34	0	0	0	0%
Moonshot v1 Moonshot AI	●○○○Low / Estimated	~45.15	0	0	0	0%
Seed-2.0-Mini ByteDance	●○○○Low / Estimated	~44.95	0	0	32	0%
Nemotron Ultra 253B NVIDIA	●○○○Low / Estimated	~44.76	0	0	0	0%
Gemini 1.0 Pro Google	●○○○Low / Estimated	~21.08	0	0	0	0%
Claude 3 Haiku Anthropic	●○○○Low / Estimated	~20.68	0	0	0	0%
LFM2-24B-A2B LiquidAI	●○○○Low / Estimated	~18.14	0	0	43	0%
Claude 4.1 Opus Thinking Anthropic	●○○○Low / Estimated	~35.58	0	0	0	0%
Ministral 3 8B (Reasoning) Mistral	●○○○Low / Estimated	~40.72	0	0	48	0%
Nova Pro Amazon	●○○○Low / Estimated	~19.56	0	0	0	0%
Ministral 3 8B Mistral	●○○○Low / Estimated	~20.29	0	0	48	0%
Mistral 7B v0.3 Mistral	●○○○Low / Estimated	~40.24	0	0	0	0%
Qwen2.5-VL-32B Alibaba	●○○○Low / Estimated	~40.19	0	0	0	0%
Llama 4 Behemoth Meta	●○○○Low / Estimated	~40.15	0	0	0	0%
LFM2.5-1.2B-Thinking LiquidAI	●○○○Low / Estimated	~15.53	0	0	43	0%
Ministral 3 3B (Reasoning) Mistral	●○○○Low / Estimated	~39.86	0	0	48	0%
MiniMax M1 80k MiniMax	●○○○Low / Estimated	~24.27	0	0	43	0%
Ministral 3 3B Mistral	●○○○Low / Estimated	~17.82	0	0	48	0%
Mistral 8x7B v0.2 Mistral	●○○○Low / Estimated	~39.47	0	0	0	0%
LFM2.5-1.2B-Instruct LiquidAI	●○○○Low / Estimated	~14.82	0	0	43	0%
Agents-A1-F16-GGUF InternScience	●○○○Low / Estimated	~60.79	0	0	0	0%
Agents-A1-FP8 InternScience	●○○○Low / Estimated	~60.79	0	0	0	0%
Agents-A1-Q4_K_M-GGUF InternScience	●○○○Low / Estimated	~60.79	0	0	0	0%
Agents-A1-Q8_0-GGUF InternScience	●○○○Low / Estimated	~60.79	0	0	0	0%
Gemma 4 E2B Google	●○○○Low / Estimated	~42.18	0	0	0	0%
Gemma 4 E4B Google	●○○○Low / Estimated	~43.25	0	0	0	0%
Sarvam 105B Sarvam	●○○○Low / Estimated	~43.14	0	0	0	0%
Sarvam 30B Sarvam	●○○○Low / Estimated	~40.91	0	0	0	0%
Mistral Medium 3 Mistral	●○○○Low / Estimated	~43.37	0	0	0	0%
Granite-4.0-350M IBM	●○○○Low / Estimated	~38.76	0	0	0	0%
Granite-4.0-H-350M IBM	●○○○Low / Estimated	~38.76	0	0	0	0%
GPT-5 nano OpenAI	●○○○Low / Estimated	~46.49	0	0	28	0%
DeepSeek R1 Distill Qwen 32B DeepSeek	●○○○Low / Estimated	~42.76	0	0	0	0%
o1-pro OpenAI	●○○○Low / Estimated	~46.07	0	0	0	0%
Exaone 4.0 1.2B LG AI Research	●○○○Low / Estimated	~39.3	0	0	0	0%
K-Exaone LG AI Research	●○○○Low / Estimated	~47.44	0	0	0	0%
Solar Pro 2 Upstage	●○○○Low / Estimated	~41.39	0	0	0	0%

Model / evidenceConfidence / score

Qwen3.7 PlusAlibaba50 sourced · 50 rankable · 0 generated

●●●●High66.02provisional score

Claude Opus 4.5Anthropic41 sourced · 41 rankable · 0 generated

●●●●High63.63provisional score

Kimi K2.5Moonshot AI40 sourced · 40 rankable · 0 generated

●●●●High58.89provisional score

Qwen3.6 PlusAlibaba39 sourced · 39 rankable · 0 generated

●●●●High64.79provisional score

Qwen3.5 397BAlibaba35 sourced · 35 rankable · 0 generated

●●●●High57.46provisional score

GLM-5Z.AI34 sourced · 34 rankable · 0 generated

●●●●High65.61provisional score

Qwen3.7 MaxAlibaba32 sourced · 32 rankable · 0 generated

●●●●High71.6provisional score

Gemini 3.5 FlashGoogle23 sourced · 23 rankable · 0 generated

●●●●High64.67provisional score

Claude Opus 5Anthropic59 sourced · 59 rankable · 0 generated

●●●○Good83.07provisional score

Qwen3.6-35B-A3BAlibaba39 sourced · 39 rankable · 0 generated

●●●○Good51.42provisional score

Qwen3.6-27BAlibaba36 sourced · 36 rankable · 0 generated

●●●○Good53.73provisional score

GPT-5.4OpenAI32 sourced · 32 rankable · 0 generated

●●●○Good73.36provisional score

Claude Opus 4.8Anthropic31 sourced · 31 rankable · 0 generated

●●●○Good76.58provisional score

Claude Opus 4.6Anthropic30 sourced · 30 rankable · 0 generated

●●●○Good67.98provisional score

GPT-5.5OpenAI29 sourced · 29 rankable · 0 generated

●●●○Good73.37provisional score

Kimi K2.6Moonshot AI29 sourced · 29 rankable · 0 generated

●●●○Good60.21provisional score

GPT-5.6 SolOpenAI23 sourced · 23 rankable · 0 generated

●●●○Good82provisional score

Muse SparkMeta22 sourced · 22 rankable · 0 generated

●●●○Good70.81provisional score

GPT-5.6 TerraOpenAI21 sourced · 21 rankable · 0 generated

●●●○Good72.91provisional score

GPT-5.6 LunaOpenAI21 sourced · 21 rankable · 0 generated

●●●○Good67.29provisional score

Gemini 3.1 ProGoogle20 sourced · 20 rankable · 0 generated

●●●○Good55.98provisional score

Claude Opus 4.7 (Adaptive)Anthropic20 sourced · 20 rankable · 0 generated

●●●○Good72.13provisional score

Claude Sonnet 4.6Anthropic18 sourced · 18 rankable · 0 generated

●●●○Good64.53provisional score

Inkling-SmallThinking Machines Lab18 sourced · 18 rankable · 0 generated

●●●○Good65.54provisional score

Nemotron 3 UltraNVIDIA17 sourced · 17 rankable · 0 generated

●●●○Good46.67provisional score

InklingThinking Machines Lab16 sourced · 16 rankable · 0 generated

●●●○Good67.04provisional score

Ling 3.0 FlashInclusionAI16 sourced · 16 rankable · 0 generated

●●●○Good53.78provisional score

GPT-5.2OpenAI14 sourced · 14 rankable · 0 generated

●●●○Good58.24provisional score

Qwen3.5-122B-A10BAlibaba13 sourced · 13 rankable · 0 generated

●●●○Good59.64provisional score

Qwen3.5-27BAlibaba13 sourced · 13 rankable · 0 generated

●●●○Good59.82provisional score

MAI-Thinking-1Microsoft13 sourced · 13 rankable · 0 generated

●●●○Good50.93provisional score

Qwen3.5-35B-A3BAlibaba13 sourced · 13 rankable · 0 generated

●●●○Good56.15provisional score

GPT-5.4 miniOpenAI12 sourced · 12 rankable · 0 generated

●●●○Good56.86provisional score

Qwen3.8 MaxAlibaba40 sourced · 9 rankable · 0 generated

●●○○Moderate79.91provisional score

Kimi K3Moonshot AI29 sourced · 9 rankable · 0 generated

●●○○Moderate80.5provisional score

DeepSeek V4 Pro 0813DeepSeek23 sourced · 23 rankable · 0 generated

●●○○Moderate61.16provisional score

Muse Spark 1.1Meta19 sourced · 19 rankable · 0 generated

●●○○Moderate76.88provisional score

Grok 4.20xAI17 sourced · 17 rankable · 0 generated

●●○○Moderate55.19provisional score

Claude Sonnet 5Anthropic16 sourced · 16 rankable · 0 generated

●●○○Moderate64.78provisional score

GLM-5.1Z.AI16 sourced · 16 rankable · 0 generated

●●○○Moderate66.91provisional score

GLM-5.2Z.AI15 sourced · 15 rankable · 0 generated

●●○○Moderate63.34provisional score

MiniMax M3MiniMax15 sourced · 15 rankable · 0 generated

●●○○Moderate68.74provisional score

Claude Mythos 5Anthropic14 sourced · 14 rankable · 0 generated

●●○○Moderate83.21provisional score

Nemotron 3 Nano Omni 30B A3BNVIDIA14 sourced · 14 rankable · 0 generated

●●○○Moderate44.4provisional score

Muse Glimmer 30BMeta13 sourced · 8 rankable · 0 generated

●●○○Moderate52.5provisional score

MiniCPM5-1BOpenBMB13 sourced · 13 rankable · 0 generated

●●○○Moderate11.36provisional score

Nemotron 3.5 Lightning 30B A3B NVFP4NVIDIA13 sourced · 13 rankable · 0 generated

●●○○Moderate26.79provisional score

Claude Fable 5Anthropic12 sourced · 12 rankable · 0 generated

●●○○Moderate82.96provisional score

Gemini 3 ProGoogle11 sourced · 11 rankable · 0 generated

●●○○Moderate67.37provisional score

GPT-5.4 nanoOpenAI11 sourced · 11 rankable · 0 generated

●●○○Moderate66.59provisional score

GPT-5.4 ProOpenAI11 sourced · 11 rankable · 0 generated

●●○○Moderate61.26provisional score

Gemma 4 12BGoogle11 sourced · 11 rankable · 0 generated

●●○○Moderate47.32provisional score

ZAYA1-8BZyphra10 sourced · 10 rankable · 0 generated

●●○○Moderate30.96provisional score

GLM-4.7Z.AI9 sourced · 9 rankable · 0 generated

●●○○Moderate60.66provisional score

Grok 4.5xAI9 sourced · 9 rankable · 0 generated

●●○○Moderate75.42provisional score

Qwen 3.6 Max (preview)Alibaba9 sourced · 9 rankable · 0 generated

●●○○Moderate59.99provisional score

Claude Sonnet 4.5Anthropic9 sourced · 9 rankable · 0 generated

●●○○Moderate54.16provisional score

Gemini 3.7 FlashGoogle8 sourced · 8 rankable · 0 generated

●●○○Moderate61.36provisional score

Kimi K2.5 (Reasoning)Moonshot AI8 sourced · 8 rankable · 0 generated

●●○○Moderate59.81provisional score

MiMo-V2.5Xiaomi8 sourced · 8 rankable · 0 generated

●●○○Moderate59.16provisional score

Hy3 PreviewTencent8 sourced · 8 rankable · 0 generated

●●○○Moderate43.96provisional score

MiniMax M2.7MiniMax16 sourced · 16 rankable · 0 generated

●○○○Low / Estimated~63.06provisional score

Step 3.7 FlashStepFun10 sourced · 10 rankable · 0 generated

●○○○Low / Estimated~50.91provisional score

GPT-5.3 CodexOpenAI7 sourced · 7 rankable · 0 generated

●○○○Low / Estimated~65.93provisional score

MiMo-V2.5-ProXiaomi7 sourced · 7 rankable · 0 generated

●○○○Low / Estimated~69.21provisional score

GPT-5.5 ProOpenAI6 sourced · 6 rankable · 0 generated

●○○○Low / Estimated~64.17provisional score

Grok 4.6xAI6 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~63.41provisional score

Agents-A1InternScience6 sourced · 6 rankable · 0 generated

●○○○Low / Estimated~60.79provisional score

Claude Opus 4.7Anthropic6 sourced · 6 rankable · 0 generated

●○○○Low / Estimated~71.87provisional score

DeepSeek V3.2DeepSeek6 sourced · 6 rankable · 0 generated

●○○○Low / Estimated~54.67provisional score

Grok 4.3xAI6 sourced · 6 rankable · 0 generated

●○○○Low / Estimated~63.95provisional score

GPT-4.1OpenAI6 sourced · 6 rankable · 0 generated

●○○○Low / Estimated~50.84provisional score

Gemma 4 31BGoogle6 sourced · 6 rankable · 0 generated

●○○○Low / Estimated~60.15provisional score

LFM2.5-8B-A1BLiquidAI6 sourced · 6 rankable · 0 generated

●○○○Low / Estimated~41.62provisional score

Gemini 3.5 Flash-LiteGoogle5 sourced · 5 rankable · 0 generated

●○○○Low / Estimated~65.29provisional score

Gemini 3 FlashGoogle5 sourced · 5 rankable · 0 generated

●○○○Low / Estimated~59.74provisional score

Claude Haiku 4.5Anthropic5 sourced · 5 rankable · 0 generated

●○○○Low / Estimated~57.06provisional score

o3-miniOpenAI5 sourced · 5 rankable · 0 generated

●○○○Low / Estimated~46.68provisional score

GPT-4.1 miniOpenAI5 sourced · 5 rankable · 0 generated

●○○○Low / Estimated~44.14provisional score

Gemini 2.5 ProGoogle5 sourced · 5 rankable · 0 generated

●○○○Low / Estimated~56.63provisional score

o1OpenAI4 sourced · 4 rankable · 0 generated

●○○○Low / Estimated~48.17provisional score

Qwen3 235B 2507Alibaba4 sourced · 4 rankable · 0 generated

●○○○Low / Estimated~56.45provisional score

Qwen3.5 PlusAlibaba4 sourced · 4 rankable · 0 generated

●○○○Low / Estimated~47.47provisional score

Gemma 4 26B A4BGoogle4 sourced · 4 rankable · 0 generated

●○○○Low / Estimated~57.16provisional score

GPT-4.1 nanoOpenAI4 sourced · 4 rankable · 0 generated

●○○○Low / Estimated~42.2provisional score

Trinity-Large-PreviewArcee AI4 sourced · 4 rankable · 0 generated

●○○○Low / Estimated~56.39provisional score

Trinity-Large-ThinkingArcee AI4 sourced · 4 rankable · 0 generated

●○○○Low / Estimated~47.71provisional score

GPT-5.1OpenAI3 sourced · 3 rankable · 0 generated

●○○○Low / Estimated~53.54provisional score

Grok 4xAI3 sourced · 3 rankable · 0 generated

●○○○Low / Estimated~59.61provisional score

GLM-4.6Z.AI3 sourced · 3 rankable · 0 generated

●○○○Low / Estimated~54.34provisional score

Ling 2.6 FlashInclusionAI3 sourced · 3 rankable · 0 generated

●○○○Low / Estimated~44.08provisional score

Command A+Cohere3 sourced · 3 rankable · 0 generated

●○○○Low / Estimated~47.56provisional score

Gemini 3.6 FlashGoogle2 sourced · 2 rankable · 0 generated

●○○○Low / Estimated~75.54provisional score

Exaone 4.0 32BLG AI Research2 sourced · 2 rankable · 0 generated

●○○○Low / Estimated~40.65provisional score

Claude 4.1 OpusAnthropic2 sourced · 2 rankable · 0 generated

●○○○Low / Estimated~44.99provisional score

Claude 4 SonnetAnthropic2 sourced · 2 rankable · 0 generated

●○○○Low / Estimated~41.97provisional score

Gemini 3.1 Flash-LiteGoogle2 sourced · 2 rankable · 0 generated

●○○○Low / Estimated~50.51provisional score

o4-mini (high)OpenAI2 sourced · 2 rankable · 0 generated

●○○○Low / Estimated~50.37provisional score

Kimi K2Moonshot AI2 sourced · 2 rankable · 0 generated

●○○○Low / Estimated~26.32provisional score

o3OpenAI2 sourced · 2 rankable · 0 generated

●○○○Low / Estimated~47.01provisional score

Qwen3 235B 2507 (Reasoning)Alibaba2 sourced · 2 rankable · 0 generated

●○○○Low / Estimated~58.47provisional score

Gemini 2.5 FlashGoogle2 sourced · 2 rankable · 0 generated

●○○○Low / Estimated~47.45provisional score

Qwen3.5 FlashAlibaba2 sourced · 2 rankable · 0 generated

●○○○Low / Estimated~47.27provisional score

Grok 3 [Beta]xAI2 sourced · 2 rankable · 0 generated

●○○○Low / Estimated~40.04provisional score

Claude 3.5 SonnetAnthropic2 sourced · 2 rankable · 0 generated

●○○○Low / Estimated~48.11provisional score

GPT-5 (high)OpenAI2 sourced · 2 rankable · 0 generated

●○○○Low / Estimated~59.08provisional score

GPT-5.2-CodexOpenAI2 sourced · 2 rankable · 0 generated

●○○○Low / Estimated~58.19provisional score

Kimi K2.7 CodeMoonshot AI2 sourced · 2 rankable · 0 generated

●○○○Low / Estimated~54.34provisional score

GPT-5.1-CodexOpenAI2 sourced · 2 rankable · 0 generated

●○○○Low / Estimated~52.77provisional score

MiMo-V2-FlashXiaomi1 sourced · 1 rankable · 0 generated

●○○○Low / Estimated~53.3provisional score

Gemini 3 Pro Deep ThinkGoogle1 sourced · 1 rankable · 0 generated

●○○○Low / Estimated~61.79provisional score

DeepSeek V3DeepSeek1 sourced · 1 rankable · 0 generated

●○○○Low / Estimated~44.28provisional score

GPT-4oOpenAI1 sourced · 1 rankable · 0 generated

●○○○Low / Estimated~40.81provisional score

Llama 4 ScoutMeta1 sourced · 1 rankable · 0 generated

●○○○Low / Estimated~39.39provisional score

Llama 4 MaverickMeta1 sourced · 1 rankable · 0 generated

●○○○Low / Estimated~22.76provisional score

Claude Opus 4.6 (Adaptive)Anthropic1 sourced · 1 rankable · 0 generated

●○○○Low / Estimated~64.72provisional score

GLM-5 (Reasoning)Z.AI1 sourced · 1 rankable · 0 generated

●○○○Low / Estimated~60.25provisional score

GLM-5-TurboZ.AI1 sourced · 1 rankable · 0 generated

●○○○Low / Estimated~66.08provisional score

MiMo-V2-ProXiaomi1 sourced · 1 rankable · 0 generated

●○○○Low / Estimated~66.94provisional score

GLM-5V-TurboZ.AI1 sourced · 1 rankable · 0 generated

●○○○Low / Estimated~62.28provisional score

Grok 4.1 Fast (Reasoning)xAI1 sourced · 1 rankable · 0 generated

●○○○Low / Estimated~59.71provisional score

MiMo-V2-OmniXiaomi1 sourced · 1 rankable · 0 generated

●○○○Low / Estimated~62.36provisional score

DeepSeek V3.2 (Thinking)DeepSeek1 sourced · 1 rankable · 0 generated

●○○○Low / Estimated~58.59provisional score

Grok 4 Fast (Reasoning)xAI1 sourced · 1 rankable · 0 generated

●○○○Low / Estimated~55.8provisional score

MiniMax M2.5MiniMax1 sourced · 1 rankable · 48 generated

●○○○Low / Estimated~58.68provisional score

GPT-OSS 120BOpenAI1 sourced · 1 rankable · 0 generated

●○○○Low / Estimated~49.23provisional score

GPT-5.1-Codex-MaxOpenAI1 sourced · 1 rankable · 0 generated

●○○○Low / Estimated~54.9provisional score

GPT-OSS 20BOpenAI1 sourced · 1 rankable · 0 generated

●○○○Low / Estimated~42.2provisional score

Nemotron 3 Super 100BNVIDIA1 sourced · 1 rankable · 0 generated

●○○○Low / Estimated~50.47provisional score

GPT-5 miniOpenAI1 sourced · 1 rankable · 60 generated

●○○○Low / Estimated~43.08provisional score

Qwen3 MaxAlibaba1 sourced · 1 rankable · 0 generated

●○○○Low / Estimated~48.26provisional score

Claude Opus 4.5 ThinkingAnthropic1 sourced · 1 rankable · 0 generated

●○○○Low / Estimated~57.46provisional score

Muse Spark 1.2Meta0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~61.67provisional score

Grok 4.1xAI0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~60.43provisional score

Hy3Tencent0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~68.02provisional score

Qwen3.5 397B (Reasoning)Alibaba0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~59.96provisional score

GPT-5.2 InstantOpenAI0 sourced · 0 rankable · 32 generated

●○○○Low / Estimated~59.47provisional score

GPT-5.2 ProOpenAI0 sourced · 0 rankable · 32 generated

●○○○Low / Estimated~66.87provisional score

GPT-5.3 InstantOpenAI0 sourced · 0 rankable · 35 generated

●○○○Low / Estimated~59.36provisional score

Grok 4.1 FastxAI0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~50.57provisional score

DeepSeek V3.1 (Reasoning)DeepSeek0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~52.72provisional score

DeepSeek V3.1DeepSeek0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~52.9provisional score

Mistral Large 3Mistral0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~49.54provisional score

GLM-4.5Z.AI0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~58.02provisional score

DeepSeek-R1DeepSeek0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~50.95provisional score

GPT-5.3-Codex-SparkOpenAI0 sourced · 0 rankable · 44 generated

●○○○Low / Estimated~57.36provisional score

Step 3.5 FlashStepFun0 sourced · 0 rankable · 50 generated

●○○○Low / Estimated~54.29provisional score

o1-previewOpenAI0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~48.42provisional score

GLM-4.5-AirZ.AI0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~46.99provisional score

GLM-4.7-FlashZ.AI0 sourced · 0 rankable · 45 generated

●○○○Low / Estimated~50.46provisional score

Gemma 3 27BGoogle0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~41.04provisional score

Mistral Small 4Mistral0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~46.31provisional score

DeepSeek LLM 2.0DeepSeek0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~54.95provisional score

Mercury 2Inception0 sourced · 0 rankable · 55 generated

●○○○Low / Estimated~48.14provisional score

GPT-5 (medium)OpenAI0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~54.29provisional score

Claude 3 OpusAnthropic0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~40.43provisional score

Nemotron 3 Nano 30BNVIDIA0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~53.36provisional score

GPT-4o miniOpenAI0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~37.22provisional score

Mistral Large 2Mistral0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~41.96provisional score

Qwen2.5-72BAlibaba0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~52.56provisional score

Llama 3.1 405BMeta0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~52.11provisional score

Nemotron 3 Super 120B A12BNVIDIA0 sourced · 0 rankable · 49 generated

●○○○Low / Estimated~51.37provisional score

Llama 3 70BMeta0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~51.22provisional score

Qwen2.5 Coder 32B InstructAlibaba0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~34.03provisional score

DeepSeek Coder 2.0DeepSeek0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~50.67provisional score

Seed 1.6ByteDance0 sourced · 0 rankable · 32 generated

●○○○Low / Estimated~50.62provisional score

Gemini 1.5 ProGoogle0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~35provisional score

Phi-4Microsoft0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~22.41provisional score

Qwen2.5-1MAlibaba0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~50.27provisional score

DeepSeekMath V2DeepSeek0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~50.27provisional score

Seed-2.0-LiteByteDance0 sourced · 0 rankable · 32 generated

●○○○Low / Estimated~50.22provisional score

Ministral 3 14B (Reasoning)Mistral0 sourced · 0 rankable · 48 generated

●○○○Low / Estimated~49.73provisional score

o3-proOpenAI0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~47.43provisional score

Ministral 3 14BMistral0 sourced · 0 rankable · 48 generated

●○○○Low / Estimated~33.89provisional score

Aion-2.0Aion Labs0 sourced · 0 rankable · 32 generated

●○○○Low / Estimated~49.04provisional score

Mixtral 8x22B Instruct v0.1Mistral0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~48.94provisional score

Grok Code Fast 1xAI0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~37.79provisional score

GPT-4 TurboOpenAI0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~26.71provisional score

Z-1Z0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~45.49provisional score

Seed 1.6 FlashByteDance0 sourced · 0 rankable · 32 generated

●○○○Low / Estimated~45.44provisional score

Nemotron-4 15BNVIDIA0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~45.44provisional score

Mistral 8x7BMistral0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~45.34provisional score

Moonshot v1Moonshot AI0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~45.15provisional score

Seed-2.0-MiniByteDance0 sourced · 0 rankable · 32 generated

●○○○Low / Estimated~44.95provisional score

Nemotron Ultra 253BNVIDIA0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~44.76provisional score

Gemini 1.0 ProGoogle0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~21.08provisional score

Claude 3 HaikuAnthropic0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~20.68provisional score

LFM2-24B-A2BLiquidAI0 sourced · 0 rankable · 43 generated

●○○○Low / Estimated~18.14provisional score

Claude 4.1 Opus ThinkingAnthropic0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~35.58provisional score

Ministral 3 8B (Reasoning)Mistral0 sourced · 0 rankable · 48 generated

●○○○Low / Estimated~40.72provisional score

Nova ProAmazon0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~19.56provisional score

Ministral 3 8BMistral0 sourced · 0 rankable · 48 generated

●○○○Low / Estimated~20.29provisional score

Mistral 7B v0.3Mistral0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~40.24provisional score

Qwen2.5-VL-32BAlibaba0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~40.19provisional score

Llama 4 BehemothMeta0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~40.15provisional score

LFM2.5-1.2B-ThinkingLiquidAI0 sourced · 0 rankable · 43 generated

●○○○Low / Estimated~15.53provisional score

Ministral 3 3B (Reasoning)Mistral0 sourced · 0 rankable · 48 generated

●○○○Low / Estimated~39.86provisional score

MiniMax M1 80kMiniMax0 sourced · 0 rankable · 43 generated

●○○○Low / Estimated~24.27provisional score

Ministral 3 3BMistral0 sourced · 0 rankable · 48 generated

●○○○Low / Estimated~17.82provisional score

Mistral 8x7B v0.2Mistral0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~39.47provisional score

LFM2.5-1.2B-InstructLiquidAI0 sourced · 0 rankable · 43 generated

●○○○Low / Estimated~14.82provisional score

Agents-A1-F16-GGUFInternScience0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~60.79provisional score

Agents-A1-FP8InternScience0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~60.79provisional score

Agents-A1-Q4_K_M-GGUFInternScience0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~60.79provisional score

Agents-A1-Q8_0-GGUFInternScience0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~60.79provisional score

Gemma 4 E2BGoogle0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~42.18provisional score

Gemma 4 E4BGoogle0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~43.25provisional score

Sarvam 105BSarvam0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~43.14provisional score

Sarvam 30BSarvam0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~40.91provisional score

Mistral Medium 3Mistral0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~43.37provisional score

Granite-4.0-350MIBM0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~38.76provisional score

Granite-4.0-H-350MIBM0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~38.76provisional score

GPT-5 nanoOpenAI0 sourced · 0 rankable · 28 generated

●○○○Low / Estimated~46.49provisional score

DeepSeek R1 Distill Qwen 32BDeepSeek0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~42.76provisional score

o1-proOpenAI0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~46.07provisional score

Exaone 4.0 1.2BLG AI Research0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~39.3provisional score

K-ExaoneLG AI Research0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~47.44provisional score

Solar Pro 2Upstage0 sourced · 0 rankable · 0 generated

●○○○Low / Estimated~41.39provisional score

Sourced = exact-source benchmark coverage. Rankable = non-generated benchmark coverage used by the provisional leaderboard. Generated = inferred from related models and excluded from ranking. Coverage = sourced share of the visible benchmark footprint.

Continue from confidence to a decision

Read the ranking methodology, inspect how leadership changed in the LLM leaderboard history, or cite the live benchmark statistics.

Frequently Asked Questions

What is benchmark confidence on BenchLM?

Score confidence (1-4 dots) indicates how much sourced benchmark data supports a model's score. A 4-dot score has at least 20 sourced rows across seven categories. A 1-dot score has limited coverage. Some benchmark-sparse models can appear provisionally when several independent public evaluation families agree.

What does "estimated" mean on BenchLM scores?

Scores marked with "Est." or "~" have limited source-backed benchmark coverage or qualify through corroborating external evaluation families. Unresolved manual values and generated rows cannot score. The verified leaderboard requires sourced benchmark support; external consensus alone never upgrades an estimated row to verified status.

How does BenchLM detect contamination risk?

BenchLM tracks two key signals: (1) benchmark provenance — whether each score is a hand-entered public row ("manual") or was generated/inferred from related data, and (2) benchmark freshness — older benchmarks that haven't been updated are more likely to have been contaminated through training data inclusion. Models with mostly generated data or stale benchmarks receive lower confidence ratings. Exact-source verification is tracked separately from this manual-vs-generated split.

What is benchmark provenance?

Provenance records where each benchmark value came from and whether the exact model, benchmark variant, harness, and score can be supported. Source-backed rows can score. Unresolved manual values and generated estimates remain in the audit trail but do not affect either public ranking view.

Which LLM benchmarks are most reliable?

Fresh, held-out benchmarks like SWE-Rebench (rolling window), Terminal-Bench 2.0, and HLE are the hardest to game. Older, saturated benchmarks like MMLU (where top models all score 97-99%) provide little signal. BenchLM weights newer, harder benchmarks more heavily and flags saturated ones as display-only.

Know when it’s worth switching models

The model to choose, the cheaper alternative, and the release we would wait on.

Read a sample issue

Join 2,000+ readers.

One email each week. Unsubscribe anytime.

Browse BenchLM

Tools & calculators

Tools

Benchmark Confidence & Contamination Flags

Confidence Distribution (Ranked Models)

How BenchLM Scores Work

Verified, provisional, and generated

Ranking Eligibility

Category Eligibility

Display-Only Benchmarks

Continue from confidence to a decision

Frequently Asked Questions

What is benchmark confidence on BenchLM?

What does "estimated" mean on BenchLM scores?

How does BenchLM detect contamination risk?

What is benchmark provenance?

Which LLM benchmarks are most reliable?

Know when it’s worth switching models