Leaked DeepSeek V4.1-Flash benchmarks top GLM-5.3 and GPT-5.6 Sol on several agent evals
QuixiAI · x · 2026-10-18
A circulating screenshot shows benchmark results for two new DeepSeek variants, DS-V4.1-Flash-1017 and DS-V4.1-Flash, compared against Kimi K3, GLM-5.3, GPT-5.6 Sol and Opus 5. Highlights: on Terminal-Bench 4.0 DS-V4.1-Flash-1017 scores 42.3, beating GLM-5.3 (37.9) and GPT-5.6 Sol (39.9), trailing only Opus 5's 51.8; it tops DeepSWE v1.1 at 77.0, the highest in the table; HLE w/o tools 42.6; Apex Bench 72.1. It still lags Opus 5 on ProgramBench (29.1) and Agents' Last Exam. Poster Quin (QuixiAI) pushes back on DeepSeek calling the update "minor," arguing the agent/coding gains are far from minor. Unofficial leak, unconfirmed.
More from Models
- Odyssey puts a fruit fly brain inside its Odyssey-3 world model, and it flies on its own — joecole · 2026-10-18
- Open models now above Sol-level, outpacing expectations as community awaits V4.1 Pro — yacineMTB · 2026-10-18
- Dev's model vibes: Opus 5.5 writes best, Sol 6.1 best value, Sonnet skippable — ricklamers · 2026-10-18
- Codex took 3 days, Opus 5.5 nailed it in an afternoon — banned user's video protest — dotey · 2026-10-18
- Yacine MTB: Users will remember how Anthropic treated them once intelligence proliferates — yacineMTB · 2026-10-18
- Mistral's Timestral forecasting models start at 5M params and run on plain CPUs — lmoroney · 2026-10-18