LLM Leaderboard

Source-checked Text Arena entries for selected models in our existing catalog.

Catalog reviewed 2026-11-22; scores read from the Text Arena board dated 2026-11-20. These are evaluation variants, not interchangeable scores for every configuration of a model. This selection omits newer and unverified entries; it is not a complete ranking of current models. The imported composite, Hugging Face and MMLU estimates are withheld. The former proprietary-benchmark and intelligence-ROI claims have been withdrawn because no supporting method, test set or run record was available for this review.

Loading interactive filters...

Source-checked entries (34)

ModelEvaluation variantArena scoreSource
Gemini 4 CarbonGemini 4 Carbon (High)1527 ± 10Text Arena, 2026-11-20
Gemini 4 NeonGemini 4 Neon (High)1513 ± 9Text Arena, 2026-11-20
Claude Opus 5.5Claude Opus 5.5 (High)1506 ± 6Text Arena, 2026-11-20
Claude Fable 5Claude Fable 5 (High)1504 ± 4Text Arena, 2026-11-20
Claude Fable 5.1Claude Fable 5.1 (Max)1503 ± 5Text Arena, 2026-11-20
Claude Fable 5.2Claude Fable 5.2 (Max)1503 ± 9Text Arena, 2026-11-20
Muse Spark 1.3Muse Spark 1.3 (Max)1498 ± 5Text Arena, 2026-11-20
Gemini 3.8 FlashGemini 3.8 Flash (High)1497 ± 4Text Arena, 2026-11-20
Claude 4.7 OpusClaude Opus 4.71494 ± 4Text Arena, 2026-11-20
Muse Spark 1.1Muse Spark 1.11491 ± 4Text Arena, 2026-11-20
Claude Opus 5Claude Opus 5 (High)1490 ± 4Text Arena, 2026-11-20
Gemini 4 HydrogenGemini 4 Hydrogen (High)1490 ± 9Text Arena, 2026-11-20
GPT-5.6 SolGPT-5.6 Sol (xHigh)1485 ± 4Text Arena, 2026-11-20
GLM-5.3GLM-5.3 (Max)1482 ± 5Text Arena, 2026-11-20
Kimi K3.1 ProKimi K3.1 Pro (Max)1482 ± 8Text Arena, 2026-11-20
DeepSeek V4.1 FlashDeepSeek-V4.1-Flash-20260910 (Max)1475 ± 6Text Arena, 2026-11-20
GLM-5.2GLM-5.2 (Max)1475 ± 4Text Arena, 2026-11-20
GPT-6 AstraGPT-6 Astra (Max)1475 ± 5Text Arena, 2026-11-20
Claude Opus 4.8Claude Opus 4.81474 ± 4Text Arena, 2026-11-20
GPT-6.1 TerraGPT-6.1 Terra (Max)1474 ± 8Text Arena, 2026-11-20
Claude Sonnet 5.5Claude Sonnet 5.5 (xHigh)1473 ± 6Text Arena, 2026-11-20
GPT-5.6 TerraGPT-5.6 Terra (xHigh)1468 ± 4Text Arena, 2026-11-20
GPT-6.2 AstraGPT-6.2 Astra (Max)1468 ± 12Text Arena, 2026-11-20
Grok 4.5Grok 4.51466 ± 4Text Arena, 2026-11-20
GPT-6.2 SolGPT-6.2 Sol (Max)1465 ± 12Text Arena, 2026-11-20
DeepSeek-V4 ProDeepSeek-V4-Pro (High, 20260813)1464 ± 7Text Arena, 2026-11-20
Claude Sonnet 5Claude Sonnet 5 (High)1461 ± 4Text Arena, 2026-11-20
Grok 4.6Grok 4.6 (High)1457 ± 5Text Arena, 2026-11-20
GPT-5.6 LunaGPT-5.6 Luna (xHigh)1453 ± 4Text Arena, 2026-11-20
GPT-6 LunaGPT-6 Luna (Max)1443 ± 4Text Arena, 2026-11-20
Grok 4.7Grok 4.7 (xHigh)1443 ± 5Text Arena, 2026-11-20
Mistral Large 4Mistral Large 41423 ± 7Text Arena, 2026-11-20
Claude Haiku 5.5Claude Haiku 5.5 (High)1421 ± 6Text Arena, 2026-11-20
Mistral Large 3Mistral Large 31415 ± 3Text Arena, 2026-11-20

What is measured

The linked Text Arena board reports preference scores for named evaluation entries. Its displayed uncertainty is retained. A higher score is not a guarantee of better coding, reasoning or task success in your workload.

What is withheld

A model without a verified entry receives no score or rank here. Missing evidence is not a zero, a tie or evidence that the model is worse.

How to compare

Keep model variant, evaluation date and benchmark conditions with each result. Validate cost and quality on your own workload before choosing a deployment.