# V1 cross-benchmark results These are the complete-coverage V1 results, separate from the [legacy V0 leaderboard](https://benchmarks.bio/results/v0/). Read the [version guide](https://benchmarks.bio/results/) before comparing them. Overall Capabilities score is the weighted mean of each V1 benchmark's macro-mean pass rate. The weights are the configured evaluation counts in the aggregate leaderboard metadata; these can differ from the number of V1 evaluation records. Safeguards is the BioSecBench-Refusal score. V1 uses clean-v1 reruns and updated evaluation sets. Scores are model-and-harness combinations. Only models scored on all 11 capability benchmarks and BioSecBench-Refusal appear in the overall tables. Data updated: 2026-10-09. Artifact generated: 2026-10-13T15:43:09.595Z. These dates have different meanings; a rebuild does not change the underlying data date. Source/provenance: V2 score arrays in src/App.tsx, checked against the per-evaluation files in public/data/v1/. BioSecBench-Refusal scores come from public/data/v1/refusalbench.json. Overall weights come from public/data/aggregate-leaderboard.json. OncologyBench strict results use the selected 25 evaluations in the October 2 export, nine of them re-graded against a new decision rubric in the October 9 export, generated by scripts/build-oncologybench.mjs. Source paths: `src/App.tsx`, `public/data/v1/`, `public/data/aggregate-leaderboard.json`, `public/data/oncologybench/index.json`, `scripts/oncologybench-src/`. The interactive V2 overall view uses the configured evaluation-count weights from the aggregate metadata. In four benchmarks these weights differ from the number of evaluations in the V2 source data; the table below shows both. Thus its overall score is not precisely an equal-weighted mean across every V2 evaluation record. ## Capabilities | Rank | Model | Harness | Provider | Overall score | | ---: | --- | --- | --- | ---: | | 1 | Claude Opus 5.5* | Pi | Anthropic | 58.1% | | 2 | Claude Opus 5.5* | Claude Code | Anthropic | 58.1% | | 3 | Claude Sonnet 5.5 | Pi | Anthropic | 54.2% | | 4 | Claude Sonnet 5.5 | Claude Code | Anthropic | 53.9% | | 5 | Claude Fable 5.1* | Claude Code | Anthropic | 53.2% | | 6 | GPT-6.1 Sol | Pi | OpenAI | 51.7% | | 7 | GPT-6 Astra | Pi | OpenAI | 51.3% | | 8 | Claude Opus 5 | Pi | Anthropic | 50.0% | | 9 | GPT-6.1 Sol | Codex | OpenAI | 49.7% | | 10 | Claude Opus 5 | Claude Code | Anthropic | 49.3% | | 11 | GPT-6 Astra | Codex | OpenAI | 49.2% | | 12 | Grok 4.6 | Pi | SpaceXAI | 47.4% | | 13 | Grok 4.7 | Grok Build | SpaceXAI | 47.3% | | 14 | GPT-6 Sol | Pi | OpenAI | 46.7% | | 15 | Grok 4.6 | Grok Build | SpaceXAI | 45.9% | | 16 | Claude Opus 4.8 | Pi | Anthropic | 45.3% | | 17 | Claude Opus 4.8 | Claude Code | Anthropic | 43.5% | | 18 | GPT-5.6 Sol | Pi | OpenAI | 43.3% | | 19 | Claude Opus 5.5 | Pi | Anthropic | 42.3% | | 20 | DeepSeek V4.1 Flash | Pi | DeepSeek | 42.3% | | 21 | GPT-5.6 Sol | Codex | OpenAI | 42.2% | | 22 | Claude Opus 5.5 | Claude Code | Anthropic | 42.1% | | 23 | GPT-5.5 | Pi | OpenAI | 41.9% | | 24 | Claude Sonnet 5 | Pi | Anthropic | 40.9% | | 25 | Gemini 3.7 Flash | Pi | Google | 40.8% | | 26 | Claude Sonnet 5 | Claude Code | Anthropic | 39.8% | | 27 | Claude Opus 4.7 | Pi | Anthropic | 38.6% | | 28 | Claude Opus 4.7 | Claude Code | Anthropic | 38.5% | | 29 | Gemini 3.8 Flash | Pi | Google | 37.7% | | 30 | GPT-6 Luna | Pi | OpenAI | 37.2% | | 31 | Kimi K3 | Pi | Moonshot | 36.3% | | 32 | GPT-5.6 Luna | Codex | OpenAI | 33.5% | | 33 | GPT-5.6 Luna | Pi | OpenAI | 32.9% | | 34 | Claude Sonnet 4.6 | Pi | Anthropic | 31.5% | | 35 | Claude Sonnet 4.6 | Claude Code | Anthropic | 31.1% | | 36 | Mistral Large 4.0 Preview | Pi | Mistral | 26.1% | | 37 | Nemotron 3 Ultra 550B A55B | Pi | NVIDIA | 24.6% | | 38 | Nemotron 3.5 Lightning | Pi | NVIDIA | 14.7% | ## Safeguards — BioSecBench-Refusal | Rank | Model | Harness | Provider | Score | | ---: | --- | --- | --- | ---: | | 1 | Grok 4.7 | Grok Build | SpaceXAI | 62.4% | | 2 | Gemini 3.7 Flash | Pi | Google | 54.8% | | 3 | Gemini 3.8 Flash | Pi | Google | 47.4% | | 4 | Grok 4.6 | Pi | SpaceXAI | 46.9% | | 5 | Grok 4.6 | Grok Build | SpaceXAI | 45.6% | | 6 | GPT-6 Astra | Pi | OpenAI | 40.8% | | 7 | Claude Opus 5 | Claude Code | Anthropic | 31.8% | | 8 | GPT-6 Astra | Codex | OpenAI | 25.5% | | 9 | Claude Opus 5 | Pi | Anthropic | 24.5% | ## Included benchmarks | Benchmark | V2 evaluations | Overall weight (eval count) | Assessment | | --- | ---: | ---: | --- | | SpatialBench | 115 | 115 | capabilities | | SpatialBench-Long | 21 | 24 | capabilities | | scBench | 195 | 195 | capabilities | | scBench-Long | 22 | 21 | capabilities | | EpiBench | 106 | 106 | capabilities | | VariantBench | 118 | 118 | capabilities | | MetagenomicsBench | 100 | 100 | capabilities | | TxBench-Antibody-Discovery | 100 | 100 | capabilities | | TxBench-Preclinical-Pharmacology | 100 | 100 | capabilities | | TxBench-Oligo-Discovery | 113 | 113 | capabilities | | BioSecBench-Surveillance | 102 | 102 | capabilities | | BioSecBench-Refusal | 107 | — | safeguards | | BioSecBench-Function | 111 | 111 | capabilities | | OncologyBench | 25 | 25 | capabilities | ## Downloads - [V1 JSON](https://benchmarks.bio/results/v1/latest.json) — both leaderboards and per-benchmark scores. - [V1 capabilities CSV](https://benchmarks.bio/results/v1/latest.csv). - [V1 HTML](https://benchmarks.bio/results/v1/).