# Science and research leaderboard

As of November 21, 2026, Claude Fable 5.2 leads our science and research leaderboard at 94.1 (Supported, $8/$40 per 1M tokens), ahead of GPT-6.2 Astra at 90.1. Gemini 4 Neon (#3, 86.3) is the lowest-cost model near the top at $3/$15 per 1M tokens; Kimi K3.1 Pro leads open-weight models at #18 (65.4). We rank 78 models (27 Supported, 51 Estimated) across 18 domain protocols, and this vertical carries zero weight in the overall score.

## Shortlist by workload and budget

- **Highest point score:** [Claude Fable 5.2](/models/claude-fable-5-2) (#1, 94.1, Supported, $8/$40 per 1M tokens) — Costs $8/$40 per 1M tokens; Gemini 4 Neon (#3, 86.3) cuts token spend to $3/$15 per 1M tokens.
- **Closest contender:** [GPT-6.2 Astra](/models/gpt-6-2-astra) (#2, 90.1, Supported, $8/$40 per 1M tokens) — Sits 4.0 points behind Claude Fable 5.2; test both on your own prompts when provider lock-in or latency matters.
- **Best API value:** [Gemini 4 Neon](/models/gemini-4-neon) (#3, 86.3, Supported, $3/$15 per 1M tokens) — Keeps 92% of the top score (86.3 vs 94.1), but gives up 7.8 points on the hardest tail cases.
- **Top open-weight model:** [Kimi K3.1 Pro](/models/kimi-3-1-pro) (#18, 65.4, Supported, Open weights · $2/$10 per 1M tokens) — Trails Claude Fable 5.2 by 28.7 points (#18 overall here), so pick it when weight control or on-prem residency outweighs peak score.

## Ranked models

| Rank | Model | Creator | Score | Conditional range | Evidence | API ($/1M in/out) | Context | Source families |
|---:|---|---|---:|---:|---|---:|---:|---:|
| #1 | [Claude Fable 5.2](/models/claude-fable-5-2) | Anthropic | 94.1 | 82.2-100.0 | Supported | $8.00 / $40.00 | 1M | 3 |
| #2 | [GPT-6.2 Astra](/models/gpt-6-2-astra) | OpenAI | 90.1 | 78.7-100.0 | Supported | $8.00 / $40.00 | 1.05M | 4 |
| #3 | [Gemini 4 Neon](/models/gemini-4-neon) | Google | 86.3 | 75.0-97.7 | Supported | $3.00 / $15.00 | 1M | 3 |
| #4 | [GPT-6 Astra](/models/gpt-6-astra) | OpenAI | 84.0 | 74.0-93.9 | Supported | $10.00 / $50.00 | 1.05M | 6 |
| #5 | [Claude Opus 5.5](/models/claude-opus-5-5) | Anthropic | 83.3 | 73.0-93.6 | Supported | $4.00 / $20.00 | 1M | 5 |
| #6 | [GPT-6.2 Sol](/models/gpt-6-2-sol) | OpenAI | 83.0 | 71.0-95.0 | Supported | $2.00 / $10.00 | 1.05M | 4 |
| #7 | [Claude Sonnet 5.5](/models/claude-sonnet-5-5) | Anthropic | 80.8 | 70.4-91.2 | Supported | $2.00 / $10.00 | 1M | 5 |
| #8 | [Gemini 4 Carbon](/models/gemini-4-carbon) | Google | 80.3 | 69.4-91.1 | Supported | $2.00 / $10.00 | 1M | 3 |
| #9 | [GPT-6.1 Sol](/models/gpt-6-1-sol) | OpenAI | 79.4 | 69.3-89.5 | Supported | $2.00 / $10.00 | 1.05M | 4 |
| #10 | [GPT-6.1 Terra](/models/gpt-6-1-terra) | OpenAI | 77.1 | 66.2-88.0 | Estimated | $1.00 / $5.00 | 1.05M | 2 |
| #11 | [Gemini 4 Argon](/models/gemini-4-argon) | Google | 73.9 | 63.0-84.7 | Supported | $2.00 / $10.00 | 1M | 4 |
| #12 | [Gemini 4 Hydrogen](/models/gemini-4-hydrogen) | Google | 73.6 | 62.7-84.5 | Estimated | $1.50 / $7.50 | 1M | 2 |
| #13 | [Claude Fable 5.1](/models/claude-fable-5-1) | Anthropic | 73.3 | 63.3-83.2 | Supported | $10.00 / $50.00 | 1M | 4 |
| #14 | [GPT-6 Sol](/models/gpt-6-sol) | OpenAI | 71.9 | 60.8-83.0 | Supported | $2.00 / $10.00 | 1.05M | 2 |
| #15 | [Claude Opus 5](/models/claude-opus-5) | Anthropic | 69.8 | 60.1-79.5 | Supported | $5.00 / $25.00 | - | 8 |
| #16 | [Claude Fable 5](/models/claude-fable) | Anthropic | 67.0 | 56.3-77.7 | Estimated | $10.00 / $50.00 | 1M+ | 3 |
| #17 | [GPT-5.6 Sol](/models/gpt-5-6-sol) | OpenAI | 66.2 | 56.5-76.0 | Supported | $4.00 / $20.00 | 1.05M | 7 |
| #18 | [Kimi K3.1 Pro](/models/kimi-3-1-pro) | Moonshot AI | 65.4 | 53.7-77.0 | Supported | $2.00 / $10.00 | 1.05M | 3 |
| #19 | [Muse Wonder](/models/muse-wonder) | Meta | 66.4 | 52.4-80.3 | Estimated | $4.00 / $20.00 | 1M | 1 |
| #20 | [Claude Opus 4.8](/models/claude-opus-4-8) | Anthropic | 63.5 | 53.3-73.6 | Supported | $5.00 / $25.00 | 1M | 5 |
| #21 | [Claude Opus 4.7](/models/claude-opus-4-7) | Anthropic | 63.2 | 49.4-77.0 | Estimated | $5.00 / $25.00 | 1M | 3 |
| #22 | [DeepSeek V4.1 Flash 1017](/models/deepseek-v4-1-flash) | DeepSeek | 62.9 | 52.7-73.0 | Supported | $0.30 / $1.20 | 1M | 4 |
| #23 | [Gemini 3.1 Pro](/models/gemini-3-1-pro) | Google | 61.6 | 46.7-76.5 | Supported | $2.00 / $12.00 | 1M | 4 |
| #24 | [Gemini 3.8 Flash](/models/gemini-3-8-flash) | Google | 61.4 | 51.5-71.4 | Supported | $0.75 / $3.75 | 1M | 7 |
| #25 | [GPT-5.5](/models/gpt-5-5) | OpenAI | 61.2 | 47.2-75.2 | Estimated | $5.00 / $30.00 | 1M | 3 |
| #26 | [Gemini 3.7 Flash](/models/gemini-3-7-flash) | Google | 61.2 | 50.5-71.8 | Supported | $0.75 / $3.75 | 1M | 4 |
| #27 | [Gemini 3.6 Flash](/models/gemini-3-6-flash) | Google | 60.9 | 47.5-74.4 | Estimated | $0.75 / $3.75 | 1M | 2 |
| #28 | [Claude Sonnet 5](/models/claude-sonnet-5) | Anthropic | 60.9 | 47.3-74.5 | Estimated | $2.00 / $10.00 | 1M | 3 |
| #29 | [Claude Opus 4.6](/models/claude-opus-4-6) | Anthropic | 60.9 | 47.1-74.7 | Estimated | $5.00 / $25.00 | 1M | 3 |
| #30 | [Claude Opus 4.6 (Adaptive)](/models/claude-opus-4-6-thinking) | Anthropic | 60.9 | 47.1-74.7 | Estimated | - | 1M | 3 |
| #31 | [Claude Haiku 5.5](/models/claude-haiku-5-5) | Anthropic | 60.8 | 50.1-71.5 | Supported | $0.10 / $0.50 | 1M | 3 |
| #32 | [GLM-5.3](/models/glm-5-3) | Z.AI | 60.7 | 50.5-70.9 | Supported | Free | 1M | 2 |
| #33 | [GPT-5.6 Terra](/models/gpt-5-6-terra) | OpenAI | 60.6 | 50.6-70.5 | Supported | $2.00 / $12.00 | 1.05M | 5 |
| #34 | [Inkling 2 Small](/models/inkling-2-small) | Thinking Machines Lab | 60.5 | 46.9-74.2 | Estimated | $0.30 / $1.00 | 1M | 2 |
| #35 | [Grok 4.6](/models/grok-4-6) | xAI | 60.0 | 50.0-70.0 | Supported | $2.00 / $6.00 | 500K | 3 |
| #36 | [Gemini 3 Pro](/models/gemini-3-pro) | Google | 61.9 | 46.1-77.5 | Estimated | $2.00 / $12.00 | 2M | 1 |
| #37 | [Grok 4.5](/models/grok-4-5) | xAI | 59.7 | 45.6-73.8 | Estimated | $2.00 / $6.00 | 500K | 2 |
| #38 | [Kimi K3](/models/kimi-k3) | Moonshot AI | 59.5 | 49.3-69.7 | Supported | $3.00 / $15.00 | 1.05M | 3 |
| #39 | [GLM-5.2](/models/glm-5-2) | Z.AI | 59.4 | 45.2-73.5 | Estimated | $1.40 / $4.40 | 1M | 3 |
| #40 | [Qwen3.7 Max](/models/qwen3-7-max) | Alibaba | 59.2 | 45.6-72.8 | Estimated | - | 1M | 4 |
| #41 | [Gemini 3.5 Flash](/models/gemini-3-5-flash) | Google | 58.8 | 45.0-72.7 | Supported | $1.50 / $9.00 | 1M | 4 |
| #42 | [GPT-5.4](/models/gpt-5-4) | OpenAI | 58.1 | 43.7-72.4 | Estimated | $2.50 / $15.00 | 1.05M | 3 |
| #43 | [GPT-5.6 Luna](/models/gpt-5-6-luna) | OpenAI | 57.6 | 47.5-67.6 | Supported | $0.20 / $1.20 | 1.05M | 4 |
| #44 | [Kimi K2.6](/models/kimi-2-6) | Moonshot AI | 55.7 | 42.0-69.4 | Estimated | $0.95 / $4.00 | 256K | 4 |
| #45 | [MiMo-V2.6-Pro](/models/mimo-v2-6-pro) | Xiaomi | 57.3 | 43.3-71.3 | Estimated | $0.43 / $0.87 | 1M | 1 |
| #46 | [o3](/models/o3) | OpenAI | 55.0 | 40.3-69.8 | Estimated | $2.00 / $8.00 | 200K | 2 |
| #47 | [Muse Spark Open](/models/muse-spark-open) | Meta | 56.4 | 40.4-72.5 | Estimated | - | 1M | 1 |
| #48 | [Mistral Large 4](/models/mistral-large-4) | Mistral | 55.1 | 42.5-67.6 | Estimated | $0.70 / $2.10 | 1M | 1 |
| #49 | [MiniMax M3](/models/minimax-m3) | MiniMax | 54.9 | 37.3-72.5 | Estimated | $0.30 / $1.20 | 1M | 1 |
| #50 | [Grok 4](/models/grok-4) | xAI | 54.1 | 37.7-70.4 | Estimated | - | 128K | 1 |
| #51 | [GPT-5](/models/gpt-5) | OpenAI | 51.9 | 36.8-67.0 | Estimated | $1.25 / $10.00 | 128K | 2 |
| #52 | [GLM-5.1](/models/glm-5-1) | Z.AI | 51.0 | 35.4-66.7 | Estimated | $1.40 / $4.40 | 203K | 3 |
| #53 | [Gemma 4 31B](/models/gemma-4-31b) | Google | 50.9 | 37.1-64.8 | Estimated | Free | 256K | 3 |
| #54 | [DeepSeek V4 Pro](/models/deepseek-v4-pro) | DeepSeek | 50.9 | 37.2-64.5 | Estimated | $1.32 / $3.96 | 1M | 2 |
| #55 | [Qwen3.6 Plus](/models/qwen3-6-plus) | Alibaba | 50.6 | 35.6-65.5 | Estimated | - | 1M | 3 |
| #56 | [o3-mini](/models/o3-mini) | OpenAI | 47.3 | 31.1-63.5 | Estimated | $1.10 / $4.40 | 200K | 2 |
| #57 | [Gemini 3.5 Flash-Lite](/models/gemini-3-5-flash-lite) | Google | 47.0 | 33.4-60.6 | Estimated | $0.30 / $2.50 | 1M | 2 |
| #58 | [DeepSeek-R1](/models/deepseek-r1) | DeepSeek | 46.9 | 29.9-63.8 | Estimated | $0.55 / $2.19 | 128K | 2 |
| #59 | [MiMo-V2.5](/models/mimo-v2-5) | Xiaomi | 48.8 | 28.8-68.9 | Estimated | - | 1M | 1 |
| #60 | [Kimi K2.5](/models/kimi-k2-5) | Moonshot AI | 46.4 | 32.0-60.9 | Estimated | $0.60 / $3.00 | 256K | 4 |
| #61 | [o1](/models/o1) | OpenAI | 45.0 | 29.6-60.3 | Estimated | $15.00 / $60.00 | 200K | 2 |
| #62 | [Qwen3.5 397B](/models/qwen3-5-397b) | Alibaba | 44.5 | 29.3-59.6 | Estimated | $0.60 / $3.60 | 128K | 3 |
| #63 | [Gemma 4 26B A4B](/models/gemma-4-26b-a4b) | Google | 42.7 | 27.8-57.8 | Estimated | Free | 256K | 2 |
| #64 | [Grok 3 Mini](/models/grok-3-mini) | xAI | 41.2 | 24.5-58.0 | Estimated | $0.30 / $0.50 | 128K | 2 |
| #65 | [MiMo-V2-Pro](/models/mimo-v2-pro) | Xiaomi | 42.3 | 21.8-62.8 | Estimated | - | 1M | 1 |
| #66 | [Gemini 3.1 Flash-Lite](/models/gemini-3-1-flash-lite) | Google | 42.3 | 24.1-60.4 | Estimated | $0.25 / $1.50 | 1M | 1 |
| #67 | [GPT-OSS 120B](/models/gpt-oss-120b) | OpenAI | 39.5 | 25.5-53.5 | Estimated | Free | 128K | 3 |
| #68 | [DeepSeek V3.1](/models/deepseek-v3-1) | DeepSeek | 39.5 | 22.1-56.9 | Estimated | Free | 128K | 2 |
| #69 | [DeepSeek V3](/models/deepseek-v3) | DeepSeek | 39.3 | 22.3-56.3 | Estimated | $0.27 / $1.10 | 128K | 2 |
| #70 | [GPT-4.1](/models/gpt-4-1) | OpenAI | 39.3 | 24.7-53.9 | Estimated | $2.00 / $8.00 | 1M | 2 |
| #71 | [Gemini 1.5 Pro](/models/gemini-1-5-pro) | Google | 37.4 | 19.6-55.1 | Estimated | $1.25 / $5.00 | 2M | 2 |
| #72 | [Claude 3.5 Sonnet](/models/claude-3-5-sonnet) | Anthropic | 35.1 | 19.0-51.1 | Estimated | $3.00 / $15.00 | 200K | 2 |
| #73 | [Grok 4.3](/models/grok-4-3) | xAI | 37.0 | 19.2-54.7 | Estimated | $1.25 / $2.50 | 1M | 1 |
| #74 | [Mistral Large 2](/models/mistral-large-2) | Mistral | 32.8 | 12.8-52.9 | Estimated | - | 128K | 2 |
| #75 | [GPT-4o](/models/gpt-4o) | OpenAI | 29.6 | 13.3-45.9 | Estimated | $2.50 / $10.00 | 128K | 2 |
| #76 | [GPT-OSS 20B](/models/gpt-oss-20b) | OpenAI | 27.9 | 12.7-43.1 | Estimated | Free | 128K | 3 |
| #77 | [GPT-4o mini](/models/gpt-4o-mini) | OpenAI | 26.2 | 9.6-42.8 | Estimated | $0.15 / $0.60 | 128K | 2 |
| #78 | [Gemma 3 27B](/models/gemma-3-27b) | Google | 23.6 | 6.3-40.8 | Estimated | Free | 32K | 2 |

## Scored domain benchmarks (18)

- [SciCode Main Standard (Kaggle)](/benchmarks/scicode) (9% ref. weight): A separate Kaggle SciCode track for main research problems without background notes.
- [FrontierPhysics](/benchmarks/frontierphysics) (8% ref. weight): FrontierPhysics, from BenchFlow, gives an AI agent a research project written by a working physicist and checks whether it gets the physics right and writes it up well enough to publish. It has 56 tasks across nine fields of physics, and each one is run three times.
- [HLE-Diamond (highest effort)](/benchmarks/hle-diamond-max) (7% ref. weight): Accuracy on the 1,000-question HLE-Diamond set without tools at each model's highest available reasoning effort.
- [HLE-Diamond](/benchmarks/hle-diamond) (7% ref. weight): Accuracy on the 1,000-question HLE-Diamond set without tools, using the benchmark owner's high-reasoning evaluation.
- [GPQA Main Few-Shot (Kaggle)](/benchmarks/gpqa) (7% ref. weight): Kaggle's few-shot run of the 448-question GPQA main set.
- [GPQA Main Zero-Shot (Kaggle)](/benchmarks/gpqa) (7% ref. weight): Kaggle's zero-shot run of the 448-question GPQA main set, kept apart from the harder Diamond subset.
- [Terminal-Bench-Science 0.1](/benchmarks/terminal-bench-science) (7% ref. weight): A benchmark of AI agents completing expert-curated research workflows across the life, physical, Earth, mathematical, and engineering sciences.
- [HLE](/benchmarks/hle) (6% ref. weight): An expert-authored benchmark designed to probe frontier knowledge and reasoning. BenchLM keeps protocol differences visible because tool-assisted and closed-book HLE runs answer different questions.
- [HLE w/o tools](/benchmarks/hlenotools) (6% ref. weight): Tool-free variant of Humanity's Last Exam that isolates a model's raw frontier reasoning.
- [SciCode](/benchmarks/scicode) (6% ref. weight): SciCode evaluates language models on generating code for realistic scientific research problems across 16 subfields of physics, math, chemistry, biology, and material science. Problems decompose into 338 subproblems requiring domain knowledge recall, scientific reasoning, and precise code synthesis. Based on real scripts from published research.
- [GPQA Diamond Zero-Shot (Kaggle)](/benchmarks/gpqa-diamond) (5% ref. weight): Kaggle's zero-shot run of the 198-question GPQA Diamond set, taken from the GPQA suite board. It stays separate from provider, Vals, Artificial Analysis, and Mercor Diamond results.
- [GPQA Diamond Few-Shot (Kaggle)](/benchmarks/gpqa-diamond) (5% ref. weight): Kaggle's few-shot run of the GPQA Diamond set at task version 5, with the October 20 run kept below it.
- [GPQA Diamond](/benchmarks/gpqa-diamond) (4% ref. weight): Graduate-level science questions from provider comparison charts and open evaluations.
- [ResearchClawBench](/benchmarks/researchclawbench) (4% ref. weight): An end-to-end autonomous scientific research benchmark with 40 tasks across 10 scientific domains, where agents receive related literature and raw data, then attempt to rediscover the hidden target paper.
- [LABBench2](/benchmarks/labbench2) (3% ref. weight): A benchmark of realistic biology-research tasks involving literature, figures, tables, databases, and bioinformatics files.
- [BioMysteryBench (human-difficult)](/benchmarks/biomysterybenchhumandifficult) (3% ref. weight): Computational biology challenges with objective answers that remained unsolved by independent human experts.
- [SciUniverse Level 1](/benchmarks/sciuniverse) (3% ref. weight): SciUniverse Level 1 tests whether models carry out scientific work across chemistry, biology, and materials science. Its 92 tasks include physical experiments, simulated lab work, and measurement analysis, with Pass@1 averaged equally across 17 task families.
- [SuperGPQA](/benchmarks/supergpqa) (1% ref. weight): An expanded version of GPQA that evaluates graduate-level knowledge and reasoning capabilities across 285 disciplines, providing comprehensive coverage of academic domains.

## Executable research code, wet-lab protocol reasoning, and graduate STEM exams

Eighteen protocols anchor the science vertical, led by SciCode Main Standard (Kaggle) (9%), FrontierPhysics (8%), HLE-Diamond (highest effort) (7%) alongside Terminal-Bench-Science, HLE-Diamond, ResearchClawBench, LABBench2, BioMysteryBench, SciUniverse, and SuperGPQA. Rather than relying on textbook recall alone, the basket pairs PhD-level physics, chemistry, and biology derivations with executable scientific computing in SciCode and containerized research workflows in Terminal-Bench-Science.

Twenty-three of the 78 ranked models hold Supported labels, while recently launched frontier models with fewer independent reruns remain visible with Estimated labels until Kaggle and academic boards complete full sweeps.

## Structured STEM benchmarks do not validate wet-lab safety or novel hypotheses

Solving closed-form quantum mechanics problems on FrontierPhysics or debugging numerical simulations in SciCode measures analytical execution on problems with known ground truth. It does not establish physical laboratory safety, reagent compatibility under real bench conditions, or statistical validity when interpreting noisy experimental assays.

Reasoning-heavy configurations also spend far more thinking tokens on HLE-Diamond and GPQA than standard chat defaults, so check effort settings and per-query latency before embedding a top-ranked model into an interactive literature assistant.

[Claude Fable 5.2 vs GPT-6.2 Astra](/compare/claude-fable-5-2-vs-gpt-6-2-astra) · [General reasoning leaderboard](/reasoning)

## Matching frontier reasoning models to simulation, literature, and budget constraints

Claude Fable 5.2 (94.1, Estimated, $8/$40 per 1M tokens) and GPT-6.2 Astra (90.1, Estimated, $8/$40 per 1M tokens) are separated by just 4.0 points at the frontier, while Claude Fable 5.2 (#1, 94.1) leads Supported rows with broader multi-evaluator confirmation. Run both top models on your own domain code or derivation set before standardizing on one provider.

For batch literature screening or parameter sweeps where $8/$40 pricing is prohibitive, Gemini 4 Neon (#3, 86.3 at $3/$15 per 1M tokens) retains 94% of the leader's score at one-quarter the token rate, and Kimi K3.1 Pro (#18, 65.4, Supported) leads self-hosted weights.

[Math leaderboard](/math) · [API price-to-performance rankings](/llm-price-performance)

## Questions

### Which LLM is best for scientific research right now?

Claude Fable 5.2 holds #1 on our scientific research leaderboard at 94.1 (Supported) as of November 21, 2026, ahead of GPT-6.2 Astra at 90.1 and Gemini 4 Neon at 86.3. Both top models hold Supported status across independent public evaluations.

### What is the best open-weight LLM for scientific research?

Kimi K3.1 Pro from Moonshot AI leads open-weight models on this scientific research board at #18 with a calibrated score of 65.4 (Supported). That trails Claude Fable 5.2 by 28.7 points, so choose Kimi K3.1 Pro when self-hosting, VPC data residency, or weight customization matters more than peak frontier accuracy.

### Which API model offers the best value for scientific research?

Gemini 4 Neon (#3, 86.3) is our top API value pick for scientific research, priced at $3/$15 per 1M tokens compared with Claude Fable 5.2 at $8/$40 per 1M tokens. You give up 7.8 calibrated points, making Gemini 4 Neon the better buy whenever those extra points do not cut downstream retries.

### Which benchmarks make up the scientific research leaderboard score?

BenchAlign v6.0 calibrates 18 open and primary domain protocols for this board, led by SciCode Main Standard (Kaggle) (9%), FrontierPhysics (8%), HLE-Diamond (highest effort) (7%), HLE-Diamond (7%). Each percentage reflects share of the vertical's reference weight inside our item-response model rather than a raw score average, and no single independent evaluator can exceed 40% when three or more evaluators contribute.

### How does the science leaderboard differ from the math and reasoning boards?

Math focuses on symbolic proof and competition problem solving, while general reasoning emphasizes domain-neutral logic and long-context deduction. This science board weights executable research software in SciCode and Terminal-Bench-Science alongside graduate physics, chemistry, and biology suites such as FrontierPhysics, GPQA Diamond, and LABBench2.

### Why do some top science models show Estimated while lower rows show Supported?

Newly released models such as Claude Fable 5.2 enter with Estimated labels when fewer independent evaluators have finished running all 18 science protocols, whereas Claude Fable 5.2 (#1) has broader multi-evaluator confirmation. Single-source estimates cannot rank above models scoring within two points below them.

## Related benchmarks and guides

- [Reasoning Leaderboard](/reasoning): General multi-step deduction and long-horizon reasoning rankings.
- [Math Leaderboard](/math): Olympiad and research-mathematics rankings on FrontierMath v2, Apex, and AIME.
- [Knowledge Leaderboard](/knowledge): Broad academic recall and subject-matter rankings.
- [Healthcare & Medical Leaderboard](/leaderboards/healthcare): Clinical reasoning and Stanford MedHELM rankings.

## Other specialist vertical leaderboards

- [Cybersecurity leaderboard](/leaderboards/cybersecurity) — #1 Gemini 4 Carbon (90.0)
- [Documents and extraction leaderboard](/leaderboards/documents) — #1 GPT-6.2 Astra (86.5)
- [Factuality and anti-hallucination leaderboard](/leaderboards/factuality) — #1 Gemini 4 Carbon (78.8)
- [Finance and accounting leaderboard](/leaderboards/finance) — #1 Gemini 4 Carbon (89.2)
- [Frontend and web development leaderboard](/leaderboards/frontend) — #1 Claude Fable 5.2 (89.5)
- [Healthcare and medical leaderboard](/leaderboards/healthcare) — #1 Claude Fable 5.2 (87.5)
- [Legal leaderboard](/leaderboards/legal) — #1 Claude Fable 5.2 (86.1)
- [Search and deep research leaderboard](/leaderboards/search) — #1 Claude Fable 5.2 (86.6)

Last updated: November 21, 2026

Canonical page: https://benchlm.ai/leaderboards/science
