TxBench launches 60-eval antibody development benchmark, top agent scores 39%
kenbwork · x · 2026-11-05
benchmarks.bio released TxBench–Antibody Development, a verifiable benchmark of 60 evaluations across seven stages of therapeutic antibody development, from candidate selection to first human dose. Tasks are graded deterministically against real experimental results — an agent passes only by recovering the original biological conclusion. Tasks include manufacturability and storage, whether process changes alter the drug, and how animal exposure and safety findings inform starting dose.
Scores are reported per model×harness since harness choice shifts results substantially. V1 leaderboard: Grok 4.7 (Grok Build) leads at 39.4%, Claude Opus 5 (Claude Code) second at 37.2%, Claude Sonnet 5.5 (Pi) third at 36.7% — no model exceeds 40%, showing how hard the task remains. arjunomics notes model capabilities are roughly orthogonal across tasks, so aggregate scores mask which models excel at disjoint skill sets.
Related event: TxBench-Antibody Launches 60-Task Benchmark Spanning Antibody Development(2 posts)→
More from Models
- Alignment lacks an engineering science: behavioral training leaves red-teaming tasks wrongly blocked — JacquesThibs · 2026-11-05
- Tencent Hunyuan Hy4 lands in OpenClaw with vision, 1M-token context, agent gains — TencentHunyuan · 2026-11-05
- Tencent Hy4 is coming to the Chutes API via a Tencent Cloud partnership — TencentHunyuan · 2026-11-05
- OpenAI API docs list a 'Using GPT-6' page, hinting at unreleased model — gabrielchua · 2026-11-05
- Researcher claims Claude sandbagging hampers safety and interp research — repligate · 2026-11-05
- Sol 6.1 hits 24.5 tps vs promised 750 tps at $60/1M output tokens — NandaVegg · 2026-11-05