DeepSeek v4 Flash Sets New Cost/Accuracy SOTA on WeirdML
teortaxesTex · x · 2026-08-03
DeepSeek v4 Flash 0731 versions achieved 57.1% and 63.0% on WeirdML, setting a new cost/accuracy SOTA ahead of the recently repriced GPT 5.6 Luna.
Although the scores seem slightly below predictions, reviewers note the model often uses submissions just for data exploration rather than pure scoring. This suggests its true agentic capabilities are likely much higher, potentially reaching GLM 5.2 levels in the right harness.
Related event: DeepSeek V4 Flash Tops OpenRouter, Costs 1% of Rivals(14 posts)→
More from Models
- OpenAI stealth-testing mysterious internal model 'Aeon' in Codex desktop app and Dots — cedric_chee · 2026-10-16
- Argon edges out astra as daily driver, but Claude still wins heavy coding work — iruletheworldmo · 2026-10-16
- NVIDIA ships FP4-quantized Qwen3.8-Flash for local use, trending on Hugging Face — nvidia · 2026-10-16
- RoboVista benchmark: Gemini-4 Argon hits 81.9% on robotics understanding — berkeley_ai · 2026-10-16
- Dev reports decision model jev beats gpt-6-luna on quality in real training pipeline — genmon · 2026-10-16
- Google won't publish Argon 4 limits in AI Pro—users compare experiences — Consistent-Text-2662 · 2026-10-16