Field guide · 2026-08-28
The top of the board is two points wide. Claude Opus 5, Fable 5, GPT-5.6 Sol, and Grok 4.6 sit inside that band. Cost, speed, context, and whether you can download the weights decide the rest.
What is the job
31 models
| Model | Index | GPQA | Term. | Agent | Speed | $/task | In/out | Ctx | ||
|---|---|---|---|---|---|---|---|---|---|---|
| Grok 4.6SpaceXAI · Closed | 61 | 91.2 | 88.4 | 58.7 | 66/s | $0.84 | $2.00 / $6.00 | 500K | ||
| Claude Opus 5Anthropic · Closed | 63 | 92.4 | 89.1 | 56.2 | 68/s | $2.34 | $5.00 / $25 | 1M | ||
| GPT-5.6 SolOpenAI · Closed | 61 | 94.6 | 88.0 | 57.8 | 62/s | $1.23 | $5.00 / $30 | 1.1M | ||
| Claude Fable 5Anthropic · Closed | 62 | 91.8 | 84.6 | 56.6 | 106/s | $3.14 | $10 / $50 | 1M | ||
| GPT-5.6 LunaOpenAI · Closed | 52 | 78.4 | 61.2 | 28.4 | 180/s | $0.05 | $0.20 / $1.20 | 1.1M | ||
| GLM-5.3Z.AI · Closed | 60 | 90.1 | 82.4 | 41.2 | 95/s | $0.68 | $1.40 / $4.40 | 1M | ||
| Kimi K3Moonshot · Other · restricted | 60 | 93.5 | 84.1 | 41.3 | 77/s | $0.84 | $3.00 / $15 | 1M | ||
| DeepSeek V4 FlashDeepSeek · MIT | 52 | 82.7 | 70.1 | 38.2 | 130/s | $0.11 | $0.44 / $1.32 | 1M | ||
| Gemini 3.7 FlashGoogle · Closed | 56 | 83.6 | 85.8 | 45.1 | 340/s | $0.40 | $0.75 / $3.75 | 1M | ||
| DeepSeek V4 ProDeepSeek · MIT | 53 | 86.8 | 78.7 | 49.6 | 78/s | $0.25 | $1.32 / $3.96 | 1M | ||
| Muse Spark 1.2Meta · Closed | 57 | 87.2 | 69.4 | 37.0 | 140/s | $0.40 | $1.25 / $4.25 | 1M | ||
| GPT-5.6 TerraOpenAI · Closed | 57 | 88.1 | 79.4 | 40.7 | 117/s | $0.51 | $2.00 / $12 | 1.1M | ||
| MiniMax M3MiniMax · MiniMax Community · restricted | 54 | 84.8 | 63.2 | 34.6 | 120/s | $0.22 | $0.30 / $1.20 | 1M | ||
| GLM-5.3 FlashZ.AI · Other · restricted | 52 | 81.4 | 66.8 | 39.1 | 140/s | $0.19 | $0.30 / $1.00 | 1M | ||
| Qwen3.8 MaxQwen · Closed | 58 | 89.4 | 75.2 | 39.7 | 47/s | $1.13 | $2.00 / $6.00 | 1M | ||
| Qwen3.8 2.4TQwen · Other · restricted | 58 | 89.1 | 74.6 | 38.9 | 55/s | $1.09 | $1.80 / $5.40 | 984K | ||
| Qwen3.8 Flash-NextQwen · Other · restricted | 50 | 80.6 | 58.4 | 37.2 | 160/s | $0.14 | $0.15 / $0.60 | 256K | ||
| Grok 4.5SpaceXAI · Closed | 56 | 86.4 | 71.2 | 44.1 | 72/s | $0.92 | $2.00 / $6.00 | 500K | ||
| Claude Sonnet 5Anthropic · Closed | 55 | 84.2 | 72.4 | 33.7 | 75/s | $1.72 | $2.00 / $10 | 1M | ||
| Mistral Large 3Mistral · Closed | 50 | 76.4 | 58.9 | 29.4 | 110/s | $0.32 | $0.50 / $1.50 | 256K | ||
| Gemini 3.5 FlashGoogle · Closed | 51 | 79.2 | 64.1 | 31.4 | 250/s | $0.48 | $0.75 / $4.50 | 1M | ||
| Grok 4 FastSpaceXAI · Closed | 48 | 72.1 | 54.8 | 24.6 | 210/s | $0.18 | $0.20 / $0.50 | 2M | ||
| Claude Opus 4.8Anthropic · Closed | 52 | 82.1 | 68.2 | 36.8 | 57/s | $2.10 | $5.00 / $25 | 1M | ||
| Gemini 3.5 Flash-LiteGoogle · Closed | 42 | 61.4 | 38.2 | 16.8 | 420/s | $0.08 | $0.15 / $1.25 | 1M | ||
| Muse GlimmerMeta · Apache-2.0 | 38 | 54.1 | 28.6 | 18.4 | 85/s | $0.04 | $0.08 / $0.24 | 128K | ||
| Nemotron 3.5 LightningNVIDIA · NVIDIA Open | 37 | 52.4 | 26.8 | 17.1 | 95/s | $0.03 | $0.05 / $0.20 | 128K | ||
| Llama 4 ScoutMeta · Llama 4 Community · restricted | 41 | 58.2 | 32.4 | 14.2 | 190/s | $0.09 | $0.11 / $0.34 | 10M | ||
| Amazon Nova 2 ProAmazon · Closed | 46 | 68.4 | 44.2 | 22.8 | 88/s | $0.45 | $0.80 / $3.20 | 1M | ||
| Ministral 3 14BMistral · Apache-2.0 | 36 | 50.0 | 22.1 | 11.4 | 70/s | $0.06 | $0.20 / $0.20 | 256K | ||
| Command ACohere · Closed | 44 | 64.8 | 29.4 | 15.6 | 92/s | $0.70 | $2.50 / $10 | 256K | ||
| Gemini 3.5 ProGoogle · Closed | — | — | — | — | — | — | $1.50 / $12 | 1M |