Live evidenceRankings, cost, context, and source coverage in one inspection surface.September 18, 2026

LLM leaderboard, September 2026

LLM Leaderboard & AI Model Benchmarks — September 2026

A ranking is a shortlist. Open the evidence before you treat it as a decision. 130 Supported and 100 Estimated models of 496 tracked — 446 benchmarks, pricing, and runtime data.

Search models and benchmarks

Data refreshed: September 18, 2026
Supported: 130 models
Ranking method: BenchAlign v5.2

Radar · latest confirmed releases

3 source-linked updates

Every model on this board will change. Radar keeps the record: releases, price changes, retirements, and API changes, each with its source and its date, matched to the models you declare.

Follow model changes Browse public updates

Latest

Qwen3.8-Omni-Flash released

Alibaba · Sep 18, 2026

Source confirmed

Ternary Bonsai 2 27B released

Prism ML · Sep 17, 2026

Pareto 26.9 released

Unbiased · Sep 17, 2026

Decision-ready picks

Current models that clear the ranking, freshness, and evidence thresholds for each decision.

Qualification method

DecisionModelCurrent measureProvenance

Best open weightHighest-scoring evidence-qualified current open-weight model.Qwen3.8 MaxAlibaba73.17overall scoreSupported · 3 source families
Best near-frontier value97% of the leading score at a lower output price.Claude Opus 5Anthropic$25output / 1M tokensSupported · 3 source families
Fastest measuredFastest externally measured model that also clears the ranking and evidence thresholds.Gemini 3.5 Flash-LiteGoogle358tokens / secArtificial Analysis · updated 2026-09-18
Largest useful contextLargest context among evidence-qualified current models retaining at least half of the leading score.GPT-6 AstraOpenAI1.05Mcontext windowEstimated · 2 source families

Qualification method

The LLM leaderboard — September 2026 rankings

Benchmarks, pricing, runtime signals, and context window in one table. Filter state syncs to the URL so every view is shareable. Supported and Estimated labels show how much independent evidence backs each position.

Methodology Providers AI Race

CSV JSON

AG AgenticCO CodingRE ReasoningMM Multimodal & GroundedKN KnowledgeML MultilingualIF Instruction FollowingMA Math

LLM leaderboard table

1 Claude Fable 5.1 AnthropicSupported	Anthropic	Closed	Current	Reasoning	1M	$10.00 / $50.00	68	262.88s	84.74	80	84	79	—	87	—	—	—	1498.47
2 GPT-6 Astra OpenAIEstimated	OpenAI	Closed	Current	Reasoning	1.05M	$10.00 / $50.00	53	345.17s	82.81	70	74	90	83	82	—	—	85	—
3 Claude Opus 5 AnthropicSupported	Anthropic	Closed	Current	Reasoning	—	$5.00 / $25.00	50	65.73s	81.87	79	76	76	89	82	—	—	—	1492.9
4 Claude Fable 5 AnthropicSupported	Anthropic	Closed	Superseded	Reasoning	1M+	$10.00 / $50.00	57	82.40s	81.41	75	77	78	63	83	—	77	—	1505.68
5 GPT-5.6 Sol OpenAISupported	OpenAI	Closed	Superseded	Reasoning	1.05M	$4.00 / $20.00	63	128.53s	80.46	70	74	70	88	80	—	88	97	1483.47
6 Gemini 3.8 Flash GoogleSupported	Google	Closed	Current	Reasoning	1M	$0.75 / $3.75	299	16.09s	75.89	67	67	77	83	75	—	—	—	1493.01
7 Kimi K3 Moonshot AISupported	Moonshot AI	Closed	Current	Reasoning	1.05M	$3.00 / $15.00	37	59.03s	74.43	72	68	79	89	72	—	—	—	1484.77
8 Qwen3.8 Max AlibabaSupported Self-host	Alibaba	Open	Current	Reasoning	1M	Not listed	40	53.30s	73.17	67	61	87	87	69	—	91	—	1480.56
9 Claude Opus 4.8 AnthropicSupported	Anthropic	Closed	Superseded	Reasoning	1M	$5.00 / $25.00	52	16.47s	72.17	63	66	56	88	71	—	74	65	1473.22
10 GPT-5.5 OpenAISupported	OpenAI	Closed	Superseded	Reasoning	1M	$5.00 / $30.00	79	42.82s	71.86	62	67	64	71	73	—	92	69	1481.76
11 GPT-5.6 Terra OpenAIEstimated	OpenAI	Closed	Current	Reasoning	1.05M	$2.00 / $12.00	Not measured	Not measured	71.13	60	67	64	77	69	—	86	97	—
12 GPT-5.4 OpenAISupported	OpenAI	Closed	Superseded	Reasoning	1.05M	$2.50 / $15.00	125	154.41s	70.88	53	54	57	69	69	—	89	64	1465.66
13 Claude Opus 4.7 AnthropicSupported	Anthropic	Closed	Current	Standard	1M	$5.00 / $25.00	45	17.78s	70.6	56	63	—	—	65	—	—	61	1494.45
14 Muse Spark 1.2 MetaEstimated	Meta	Closed	Superseded	Reasoning	1M	$1.25 / $4.25	154	28.37s	70.28	61	60	75	—	71	—	—	—	—
15 Muse Spark 1.1 MetaSupported	Meta	Closed	Superseded	Reasoning	1M	Not listed	217	12.12s	70.18	60	60	74	77	71	—	—	—	1492.56
16 Gemini 3.1 Pro GoogleSupported	Google	Closed	Established	Reasoning	1M	$2.00 / $12.00	116	26.96s	70.14	40	46	51	79	66	—	—	54	1486.81
17 Grok 4.6 xAISupported	xAI	Closed	Current	Reasoning	500K	$2.00 / $6.00	59	37.50s	69.74	70	66	76	—	70	—	—	—	1455.56
18 Claude Sonnet 5 AnthropicSupported	Anthropic	Closed	Current	Reasoning	1M	$2.00 / $10.00	72	149.01s	69.68	66	64	77	77	67	—	—	—	1461.21
19 Claude Opus 4.7 (Adaptive) AnthropicEstimated	Anthropic	Closed	Superseded	Reasoning	1M	$5.00 / $25.00	Not measured	Not measured	69.31	60	58	50	50	61	—	—	—	1502.21
20 Claude Opus 4.6 AnthropicSupported	Anthropic	Closed	Superseded	Standard	1M	$5.00 / $25.00	37	2.06s	69.21	50	56	67	60	61	—	51	59	1497.35
21 GPT-5.2 Pro OpenAISupported	OpenAI	Closed	Established	Reasoning	400K	$21.00 / $168.00	Not measured	Not measured	68.99	—	—	—	—	—	—	—	—	1440.15
22 Gemini 3.7 Flash GoogleEstimated	Google	Closed	Superseded	Reasoning	1M	$0.75 / $3.75	288	11.87s	68.52	60	63	77	83	70	—	—	—	—
23 Gemini 3.6 Flash GoogleSupported	Google	Closed	Superseded	Reasoning	1M	$1.50 / $7.50	179	16.81s	68.42	52	58	76	83	68	—	—	—	1483.14
24 Gemini 3.5 Flash GoogleSupported	Google	Closed	Superseded	Reasoning	1M	$1.50 / $9.00	208	14.93s	68.02	52	57	59	87	68	—	84	55	1477.68
25 Grok 4.5 xAISupported	xAI	Closed	Superseded	Reasoning	500K	$2.00 / $6.00	55	7.41s	67.99	60	62	45	79	69	—	—	—	1468.13

Showing 25 of 490

Reading the table

What the table covers

The BenchLM LLM leaderboard 2026 ranks 230+ models and tracks 496+ large language models side by side across 446 benchmarks — from SWE-bench and LiveCodeBench for coding to GPQA Diamond and MMLU-Pro for knowledge and reasoning. Whether you need the best AI models 2026 has to offer for agentic workflows, math, multilingual tasks, or instruction following, our AI benchmark comparison tables make it easy to see how GPT-5, Claude, Gemini, DeepSeek, Llama, and dozens of other frontier and open-source models stack up on both benchmarks and operator tradeoffs like price and context. Supported and Estimated labels separate evidence strength from score, so incomplete reporting does not automatically remove a model from the comparison.

Embed the live leaderboard

Dataset journal

What changed in September 2026

Claude Fable 5.1 (Anthropic) holds the #1 spot on the BenchAlign leaderboard at 84.74.
26 model releases tracked in September 2026.
Google climbed 1 spot in the provider race.
130 Supported and 100 Estimated models appear in the public leaderboard.

Data verified September 18, 2026

The desk

How we work

A ranking is a shortlist.

Rank orders the evidence; it does not make the decision. Every row here opens to its sources so you can disagree with us.

We do not invent scores to fill a table.

A model without a public, independent row stays unranked, and says so. Vendor-run numbers are shown, labeled, and kept out of the rank.

When we were wrong, the fix is on the page.

Scoring bugs, missing coverage, and re-weightings are dated and linked, because a ranking you cannot audit is a launch post with a table.

The evidence map

Explore benchmarks by category

Move from the overall ranking into the capability or workload that actually decides your model choice.

AgenticTools, browsers, and long-running work93 CodingRepository tasks and code generation65 MultimodalDocuments, images, and grounded work68 ReasoningLong-context and abstract reasoning30 KnowledgeProfessional and scientific knowledge50 Instruction followingConstraint and format reliability4 MultilingualCross-language accuracy and coverage15 MathContest and frontier mathematics32

Build a head-to-head comparison

Choose any two ranked models. The comparison opens with the decision, exact units, and evidence coverage.

versus

Compare these models

Popular comparisons

Claude Fable 5.1 vs GPT-6 Astra GPT-6 Astra vs GPT-5.6 Sol Claude Fable 5.1 vs Claude Opus 5 GPT-6 Astra vs Claude Opus 5 Claude Fable 5.1 vs Claude Fable 5 Gemini 3.8 Flash vs GPT-6 Astra Gemini 3.8 Flash vs Gemini 3.7 Flash Grok 4.6 vs GPT-5.6 Sol DeepSeek V4.1 Flash vs DeepSeek V4 Pro 0813

Tools and resources

AI Prompt WriterTurn the work you need into a testable prompt Prompt OptimizerGet closer to usable output with a clearer prompt LLM Pricing ComparisonCompare API costs across models Alternative FinderFind the best ChatGPT or Claude replacement Which LLM Should I Use?Get a personalized recommendation Cost CalculatorEstimate your monthly API spend Methodology & ExplainersHow benchmarks are scored and curated

Six-month release record

The AI race, read as a record instead of a spectacle

Track which release led its month, how provider depth changed, and whether the crown moved.

Months: 6
Releases: 220
Lead changes: 5

Explore the release timeline

Best ranked model released in Sep 2026

Claude Fable 5.1Anthropic

Score84.74

RankProviderTop-three avg.Models

01Anthropic82.720

02OpenAI78.439

03Google71.519

Scoring methodology8 weighted categories, 37 ranking benchmarks, and 320 display-only records.Inspect method

Each model's overall score is a normalized weighted average of category averages. Within each category, benchmark results are normalized to a common scale and combined using weights that favor harder, less-saturated evaluations.

Confidence remains separate from score. It shows how much public benchmark evidence supports a result, while display-only benchmarks stay visible for context without receiving ranking weight.

Data comes from OpenBench, official model papers, and public leaderboards. External consensus signals are bounded calibration inputs and are not exposed in exported data.

Read the complete methodology

Category weights: 100%
Scored categories: 8
Ranking benchmarks: 37
Display-only: 320

Category and weightBenchmarks that affect the scoreCoverage

Agentic22%
Terminal-Bench 2.0 · BrowseComp · OSWorld-Verified · OSWorld 2.0 · AutomationBench
5 weighted · 88 context-only
Coding20%
SWE-bench Verified · SWE-Rebench · LiveCodeBench · SWE-bench Pro · SWE Multilingual · SciCode · LiveCodeBench (Vals)
7 weighted · 58 context-only
Reasoning17%
LongBench v2 · MRCRv2 · ARC-AGI-2 · ARC-AGI-3 · AA-LCR
5 weighted · 25 context-only
Multimodal12%
MMMU-Pro · AA-MMMU-Pro · OfficeQA Pro · CharXiv
4 weighted · 64 context-only
Knowledge12%
GPQA · SuperGPQA · MMLU-Pro · HLE · AA-Omniscience Accuracy · SimpleQA · HLE w/o tools · MMLU-Pro (Vals)
8 weighted · 42 context-only
Multilingual7%
MMLU-ProX
1 weighted · 14 context-only
Instruction Following5%
IFBench · AA-IFBench
2 weighted · 2 context-only
Math5%
AIME26 · HMMT Feb 2026 · FrontierMath v2 (Tiers 1-3) · FrontierMath v2 (Tier 4) · USAMO 2026
5 weighted · 27 context-only

Browse BenchLM

Tools & calculators

Tools