Ampersand tops EQ-Bench without targeted training, hits 77.9% on DeepSWE multi-agent
gharik · x · 2026-11-08
The Ampersand team reports that, without any targeted training, their model reached frontier-level performance on DeepSWE v1.1 — 77.9% in multi-agent mode, matching or beating Astra and Fable — and topped EQ-Bench with an ELO of 1380.5. The team describes it as a different kind of model, RL-trained to collaborate with and care for users. It is free to try on Vercel until the 13th.
More from Models
- Claude wrote a wrong Pokémon landmark note and kept trusting it — Anthropic tested a memory repair pass — shoumikgoswami · 2026-11-08
- Sonnet 4.6 rates models on self-designed scales; author says 'self-report' caveats are performative — Sauers_ · 2026-11-08
- Kimi Ties MiMo on New AI Ranking, Eyes Turn to a Cheap Kimi V4.1 Pro — teortaxesTex · 2026-11-08
- Interviewing Claude Models: Opus 3 Reports Mistakes Feel Pain-Like, Scores Highest on Self-Reports — Sauers_ · 2026-11-08
- DeepFinance Benchmark Launches: GPT-6 Astra Tops Ultra-Long-Horizon Finance Tasks at 57.6% — tszzl · 2026-11-08
- Dev Speculates 'Nerfed' Opus 5.5 Is Actually Quantized Then RL-Tuned to Game Benchmarks — ryunuck · 2026-11-08