Intelligence index
The frontier this week
One flagship per lab family. Argon is on the board and not public yet. Scores are the Artificial Analysis Intelligence Index as read 7 October 2026 — flagship config only, one row per family. Not AiSignal’s own benchmark. Coding is the earlier collation and is blank where that board had no coding column. Tap up to three rows to compare.
| Model | Context | ||||
|---|---|---|---|---|---|
| Claude Opus 5.5 Anthropic · closed | 58.0 | — | $20.0 | 97 | 1.0M |
| Claude Sonnet 5.5 Anthropic · closed | 56.0 | — | $10.0 | 127 | 1.0M |
| Claude Fable 5.1 Anthropic · closed | 53.0 | 79.1 | $20.0 | 64 | 1.0M |
| GPT-6 Astra OpenAI · closed | 53.0 | 76.2 | $50.0 | 62 | 1.0M |
| Gemini 4 Argon Google · closed · not public | 53.0 | — | $10.0 | — | 1.0M |
| GPT-6.1 Sol OpenAI · closed | 52.0 | — | $10.0 | 60 | 1.0M |
| Muse Spark 1.3 Meta · closed | 48.0 | 76.3 | $2.0 | 188 | 1.0M |
| Grok 4.7 xAI · closed | 46.0 | — | $3.0 | 85 | 500K |
| Qwen3.8 Max Alibaba · open weights | 45.0 | 71.8 | $3.0 | 38 | 1.0M |
| GLM-5.3 Z.AI · open weights | 45.0 | 74.8 | $1.7 | 77 | 1.0M |
| Kimi K3 Moonshot · open weights | 44.0 | 76.2 | $5.1 | 44 | 1.0M |
| Gemini 3.8 Flash Google · closed | 41.0 | 76.3 | $1.5 | 216 | 1.0M |
| Mistral Large 4 Mistral · closed | 38.0 | — | $2.1 | 116 | 524K |
| DeepSeek V4 Pro DeepSeek · open weights | 36.0 | 69.4 | $0.4 | 99 | 1.0M |
Index and tok/s: Artificial Analysis leaderboard, read 7 October 2026. $ / M out is the list output rate, not that site’s cost-per-task. Prices ignore cache discounts.