Kimi K3.1 Pro tops open-weight models at 77% on LongTail-SWE; DeepSeek V4.1 Flash hits 76%
Fabulous_Pollution10 · reddit · 2026-11-19
New results on LongTail-SWE (202 real GitHub tasks in 13 less-common languages):
- Kimi K3.1 Pro: 77%, the top open-weight model, burning 5.6M tokens per attempt
- DeepSeek V4.1 Flash 1017: 76%, up sharply from 57% for V4 Flash
- Kimi K3 re-run with 1M context (vs 262K) scores 72%, +8 points
Per the leaderboard chart, closed models Opus 5.5 and GPT-6.1 Sol lead at 82% each; open-weight GLM-5.3 Flash (48%) and Qwen3.8 27B (43%) still trail far behind. Some task tests were fixed in this v1.2 run; full board at longtail-swe.com.
More from Models
- Heavy testing finds Astra 6.2 High slightly worse than 6.0 and 2x slower — IridiumEagle · 2026-11-19
- Heavy Astra 6.2 testing still can't exhaust the $100 plan's usage quota — prd_008 · 2026-11-19
- Gemini 4 Carbon Deciphers an Unread 1466 Sforza Renaissance Cipher Unaided — fofrAI · 2026-11-19
- Gemini's mysterious "Hydrogen" selector appears to route to Gemini 3.1 Pro — Neestor00 · 2026-11-19
- Fable 5.2 vs GPT-6.2 Astra: 108M vs 24M tokens to build the same PS1-style game — heypearlai · 2026-11-19
- Same game prompt: Astra ships for $17 vs Fable's $65 — 3x cheaper, fewer bugs — heypearlai · 2026-11-19