GPT-6.1 Sol reshapes WebDev Pareto frontier at blended $8/M tokens, ranks No.2 on Code Arena
arena · x · 2026-10-01
Code Arena's WebDev leaderboard (830,738 votes, 136 models) has a new Pareto-frontier entrant: OpenAI's GPT-6.1 Sol, at a blended $8/M tokens ($2 input / $10 output), scores 1759 — second only to Anthropic's claude-opus-5.5-max (1818, $16/M) while costing half as much.
Other Pareto-optimal models on the board:
- qwen3.8-max (Alibaba): 1671, $4.23/M
- muse-spark-1.3-max (Meta): 1655, $3.50/M
- qwen3.8-flash-next: 1638, $0.39/M
- glm-5.3-flash (Z.ai, MIT license): 1615, just $0.17/M
- granite-4.1-8b (IBM, Apache 2.0): 1191, $0.09/M
Notably, open-source GLM-5.3-flash lands within 200 points of frontier closed models at 1/100th the price.
More from Models
- Rumor: Gemini 4.0 Pro, GLM 5.5 and more new models expected within 4-6 weeks — bindureddy · 2026-10-01
- MMBU Challenge details: three tracks, API evals for 27B+ models, adaptation scored by lift — LiorOnAI · 2026-10-01
- Lior and Stanford launch $100K MMBU Challenge to test if biomedical AI really sees — LiorOnAI · 2026-10-01
- Zero-shot classification model Lev trends on Hugging Face — interfaze-ai · 2026-10-01
- Unverified Reddit post claims GPT-6.1 Sol lands third on Humanity's Last Exam Diamond — 141_1337 · 2026-10-01
- True Positive Weekly #180: Xiaomi's MIT-licensed MiMo-V2.6, physicist-style LLM pruning, OpenHands — burkov · 2026-10-01