Dashboard

The state of language models

Rankings, releases and benchmark movement · updated September 2026
Models tracked
568
86 organizations
Open weights
375
66% of catalog
Releases · 90 days
48
tracked in the last quarter
Open–closed gap
275.5 Elo
GLM-5.3 leads open
Quality vs. price
Frontier Elo against output price, log scale
Pareto frontier
050010001500200025003000$0.1$1$10$100Pareto frontier — best Frontier Elo at each priceClaude Fable 5.1Claude Opus 5GLM-5.3GPT-6 AstraMuse Spark 1.3Qwen3.8-Max-0902
Latest releases
Model explorer →
Sep 2026GPT-6 AstraOpenAI3305.4CLOSEDSep 2026Gemini 3.8 FlashGoogle2949.4CLOSEDSep 2026Gemini 3.8 Flash CyberGoogle1000.0CLOSEDSep 2026Muse Spark 1.3Meta3269.8CLOSEDSep 2026Qwen3.8-Max-0902Alibaba3092.0CLOSEDSep 2026Claude Fable 5.1Anthropic3144.2CLOSEDSep 2026Claude Mythos 5.1Anthropic3523.3CLOSEDAug 2026GLM-5.3-FlashZhipu AI / Z.ai (Tsinghua)3017.6OPEN
Top ranked
All rankings →
1GPT-6 Astra3305.4
2Muse Spark 1.33269.8
3Claude Fable 5.13144.2
4Qwen3.8-Max-09023092.0
5Claude Opus 53078.5
6GLM-5.33029.9
7GLM-5.3-Flash3017.6
8Kimi K33005.1
Biggest movers
Elo gain over previous family release
Sarvam-105Bvs Sarvam-1 (2B)+1629.1SmolLM3 3B (Think)vs SmolLM2 1.7B+1004.8Hy3vs Hunyuan-A13B+946.4SmolLM3 3B (No Thinking)vs SmolLM2 1.7B+812Phi-4-reasoningvs Phi-4-mini (3.8B)+756.8
Quick compare