# TxBench-Oligo-Discovery results > Agentic benchmark tasks for oligonucleotide discovery and development decisions. Data updated: 2026-09-20 Artifact generated: 2026-09-21T21:43:49.599Z - Domain: Therapeutics - Horizon: short - Assessment: Capabilities - Published score scope: Full benchmark (113 evaluations) Scores use a 0–100 percentage scale. Consult the repository and paper for benchmark-specific task construction, grading, aggregation, and uncertainty methodology. Results compare model-and-harness combinations. | Rank | Model | Harness | Provider | Score | | ---: | --- | --- | --- | ---: | | 1 | GPT-6 Astra | Codex | OpenAI | 55.5% | | 2 | GPT-6 Astra | Pi | OpenAI | 53.4% | | 3 | Claude Opus 5 | Claude Code | Anthropic | 48.1% | | 4 | Grok 4.6 | Pi | SpaceXAI | 44.0% | | 5 | Grok 4.6 | Grok Build | SpaceXAI | 42.8% | | 6 | Claude Opus 5 | Pi | Anthropic | 42.2% | | 7 | Claude Opus 4.8 | Claude Code | Anthropic | 39.9% | | 8 | Claude Opus 4.8 | Pi | Anthropic | 39.4% | | 9 | Claude Sonnet 5 | Claude Code | Anthropic | 39.3% | | 10 | Gemini 3.5 Flash | Pi | Google | 38.6% | | 11 | Grok 4.5 | Pi | SpaceXAI | 38.4% | | 12 | GPT-5.6 Sol | Codex | OpenAI | 34.5% | | 13 | DeepSeek V4.1 Flash | Pi | DeepSeek | 33.9% | | 14 | GPT-5.5 | Pi | OpenAI | 33.6% | | 15 | GPT-5.6 Terra | Pi | OpenAI | 33.3% | | 16 | GPT-5.6 Terra | Codex | OpenAI | 32.5% | | 17 | GPT-5.5 | Codex | OpenAI | 30.4% | | 18 | GPT-5.6 Luna | Codex | OpenAI | 23.9% | | 19 | GPT-5.6 Luna | Pi | OpenAI | 23.3% | ## Sources and downloads - [Paper PDF](/papers/txbench-oligo.pdf) - [Original publication](https://latch.bio/txbench-oligo) - [Repository](https://github.com/latchbio/txbench-oligo-discovery) - [Interactive benchmark](https://benchmarks.bio/txbench-oligo) - [JSON](https://benchmarks.bio/benchmarks/txbench-od/latest.json) - [CSV](https://benchmarks.bio/benchmarks/txbench-od/latest.csv) - [V1 cross-benchmark leaderboard](https://benchmarks.bio/results/v1/)