The highest-scoring models on the Artificial Analysis Intelligence Index at each effort level. Each chart shows a model at the best setting it has at or below that level: a model with no variant at the level appears at its highest setting below it, so a model that only has High is on the max, xhigh and high charts. A model with no labelled effort setting could be anything up to max, so it appears only on the max chart, marked with a dagger. Columns use the color Artificial Analysis gives each lab. Data last changed 2026-11-05 06:55 UTC.
The thin bar beside a column is what one full run of the index costs on that lab's subscription, read off the dollar axis on the right. That axis is shared by all four charts, so the bars show how cost falls as effort drops, and it is a different measure from the score, so it has its own scale. It is Artificial Analysis's cost for the run divided by how much more usage the plan gives than API list prices (Anthropic: Claude Max 20x 45.3x, OpenAI: ChatGPT 10.25x, SpaceXAI: SuperGrok + X Premium+ 80x), as measured by Paweł Huryn. Only labs with a measured multiplier get a bar. These are one person's estimates on one model per vendor, so treat the bars as rough. The table views give the exact figures.
Charts for: max · xhigh · high · medium · best model one level down · coding agents. Links at: xhigh · high · medium · bookmark
| # | Model | Lab | Setting | Index | Full index run: API → subscription |
|---|---|---|---|---|---|
| 1 | Gemini 4 Carbon (High) | high | 60.0 | – | |
| 2 | Claude Opus 5.5 (Max) | Anthropic | max | 59.1 | $8,910 → $197 on Claude Max 20x |
| 3 | GPT-6 Astra (Max) | OpenAI | max | 56.9 | $5,661 → $552 on ChatGPT |
| 4 | GPT-6.1 Sol (Max) | OpenAI | max | 54.6 | $1,124 → $110 on ChatGPT |
| 5 | Claude Sonnet 5.5 (Max) | Anthropic | max | 54.5 | $7,981 → $176 on Claude Max 20x |
| 6 | GPT-6.1 Terra (Max) | OpenAI | max | 51.3 | $972 → $95 on ChatGPT |
| 7 | Gemini 4 Argon (High) | high | 50.4 | – | |
| 8 | Muse Wonder Full (Max) | Meta | max | 50.1 | – |
| 9 | Claude Fable 5.1 (Max) | Anthropic | max | 49.2 | $13,327 → $294 on Claude Max 20x |
| 10 | Kimi K3.1 Pro | Kimi | not labelled | 46.5 | – |
| # | Model | Lab | Setting | Index | Full index run: API → subscription |
|---|---|---|---|---|---|
| 1 | Gemini 4 Carbon (High) | high | 60.0 | – | |
| 2 | Claude Opus 5.5 (Xhigh) | Anthropic | xhigh | 57.4 | $4,766 → $105 on Claude Max 20x |
| 3 | GPT-6 Astra (Xhigh) | OpenAI | xhigh | 55.3 | $3,746 → $365 on ChatGPT |
| 4 | GPT-6.1 Sol (Xhigh) | OpenAI | xhigh | 52.5 | $718 → $70 on ChatGPT |
| 5 | Gemini 4 Argon (High) | high | 50.4 | – | |
| 6 | Claude Fable 5.1 (Xhigh) | Anthropic | xhigh | 50.0 | $10,097 → $223 on Claude Max 20x |
| 7 | Claude Sonnet 5.5 (Xhigh) | Anthropic | xhigh | 49.9 | $2,902 → $64 on Claude Max 20x |
| 8 | GPT-6.1 Terra (Xhigh) | OpenAI | xhigh | 45.3 | $400 → $39 on ChatGPT |
| 9 | Muse Wonder (High) | Meta | high | 40.5 | – |
| 10 | Grok 4.7 (Xhigh) | SpaceXAI | xhigh | 35.5 | $5,784 → $72 on SuperGrok + X Premium+ |
| # | Model | Lab | Setting | Index | Full index run: API → subscription |
|---|---|---|---|---|---|
| 1 | Gemini 4 Carbon (High) | high | 60.0 | – | |
| 2 | GPT-6 Astra (High) | OpenAI | high | 54.8 | $3,204 → $313 on ChatGPT |
| 3 | Claude Opus 5.5 (High) | Anthropic | high | 51.8 | $2,716 → $60 on Claude Max 20x |
| 4 | Gemini 4 Argon (High) | high | 50.4 | – | |
| 5 | GPT-6.1 Sol (High) | OpenAI | high | 50.3 | $573 → $56 on ChatGPT |
| 6 | Claude Fable 5.1 (High) | Anthropic | high | 47.2 | $6,618 → $146 on Claude Max 20x |
| 7 | GPT-6.1 Terra (High) | OpenAI | high | 42.4 | $267 → $26 on ChatGPT |
| 8 | Claude Sonnet 5.5 (High) | Anthropic | high | 41.5 | $1,268 → $28 on Claude Max 20x |
| 9 | Muse Wonder (High) | Meta | high | 40.5 | – |
| 10 | Grok 4.7 (High) | SpaceXAI | high | 35.2 | $4,722 → $59 on SuperGrok + X Premium+ |
| # | Model | Lab | Setting | Index | Full index run: API → subscription |
|---|---|---|---|---|---|
| 1 | Gemini 4 Carbon (Medium) | medium | 58.3 | – | |
| 2 | GPT-6 Astra (Medium) | OpenAI | medium | 51.6 | $2,638 → $257 on ChatGPT |
| 3 | Claude Opus 5.5 (Medium) | Anthropic | medium | 48.2 | $2,031 → $45 on Claude Max 20x |
| 4 | GPT-6.1 Sol (Medium) | OpenAI | medium | 46.3 | $389 → $38 on ChatGPT |
| 5 | Claude Fable 5.1 (Medium) | Anthropic | medium | 43.2 | $4,897 → $108 on Claude Max 20x |
| 6 | GPT-6.1 Terra (Medium) | OpenAI | medium | 38.2 | $190 → $19 on ChatGPT |
| 7 | Grok 4.7 (Low) | SpaceXAI | low | 31.1 | $2,169 → $27 on SuperGrok + X Premium+ |
| 8 | Claude Sonnet 5.5 (Medium) | Anthropic | medium | 30.8 | $679 → $15 on Claude Max 20x |
| 9 | Gemini 3.8 Flash (Medium) | medium | 30.1 | – | |
| 10 | Claude Haiku 5.5 (Medium) | Anthropic | medium | 24.6 | $73 → $1.60 on Claude Max 20x |
For each lab, its best model one effort level below its next model, for example Opus 5.5 at high against Sonnet 5.5 at xhigh. A lab appears when it has two models with matching effort levels. The slate bar is Artificial Analysis's end-to-end response time, labelled in seconds: the time to the first answer token (thinking included) plus a 500-token answer at the model's speed, so lower keeps you in the flow. The table view also gives the first-token time and speed separately.
| Lab | Levels | Best model | Next model | Index difference |
|---|---|---|---|---|
| medium vs high | Gemini 4 Carbon · medium: 58.3 | Gemini 4 Argon · high: 50.4 first answer token 40.5 s, 83 tokens/s, end-to-end 46.5 s | +7.9 | |
| Anthropic | medium vs high | Claude Opus 5.5 · medium: 48.2 · $2,031 → $45 on Claude Max 20x first answer token 16.4 s, 80 tokens/s, end-to-end 22.7 s | Claude Sonnet 5.5 · high: 41.5 · $1,268 → $28 on Claude Max 20x first answer token 10.9 s, 103 tokens/s, end-to-end 15.8 s | +6.7 |
| Anthropic | high vs xhigh | Claude Opus 5.5 · high: 51.8 · $2,716 → $60 on Claude Max 20x first answer token 34.5 s, 80 tokens/s, end-to-end 40.8 s | Claude Sonnet 5.5 · xhigh: 49.9 · $2,902 → $64 on Claude Max 20x first answer token 29.7 s, 107 tokens/s, end-to-end 34.3 s | +1.9 |
| Anthropic | xhigh vs max | Claude Opus 5.5 · xhigh: 57.4 · $4,766 → $105 on Claude Max 20x first answer token 127.2 s, 85 tokens/s, end-to-end 133.1 s | Claude Sonnet 5.5 · max: 54.5 · $7,981 → $176 on Claude Max 20x first answer token 463.0 s, 139 tokens/s, end-to-end 466.6 s | +2.9 |
| OpenAI | medium vs high | GPT-6 Astra · medium: 51.6 · $2,638 → $257 on ChatGPT first answer token 5.1 s, 57 tokens/s, end-to-end 13.9 s | GPT-6.1 Sol · high: 50.3 · $573 → $56 on ChatGPT first answer token 44.4 s, 60 tokens/s, end-to-end 52.7 s | +1.3 |
| OpenAI | high vs xhigh | GPT-6 Astra · high: 54.8 · $3,204 → $313 on ChatGPT first answer token 48.1 s, 63 tokens/s, end-to-end 56.1 s | GPT-6.1 Sol · xhigh: 52.5 · $718 → $70 on ChatGPT first answer token 126.1 s, 71 tokens/s, end-to-end 133.1 s | +2.3 |
| OpenAI | xhigh vs max | GPT-6 Astra · xhigh: 55.3 · $3,746 → $365 on ChatGPT first answer token 149.4 s, 66 tokens/s, end-to-end 157.0 s | GPT-6.1 Sol · max: 54.6 · $1,124 → $110 on ChatGPT first answer token 315.1 s, 79 tokens/s, end-to-end 321.4 s | +0.7 |
Artificial Analysis Coding Agent Index v2.0 (higher is better): composite of FrontierCode v1.1, Terminal-Bench 4.0 and SWE-Atlas-QnA-AA, each benchmark equally weighted. Cost is average pay-per-token API cost per task; many harnesses are used through subscriptions instead. Data last changed 2026-10-29 18:31 UTC. Source: artificialanalysis.ai/agents/coding-agents.
Highlighted defaults plus Muse Spark variants for the cheap-vs-best read: Muse Spark 1.3 (max) is the cheap Meta API fallback, Muse Wonder Full / Wonder are the best Meta scores.
| # | Agent | Lab | Index | Cost / task (API) | Time / task |
|---|---|---|---|---|---|
| 1 | Claude Code - Claude Opus 5.5 (Max, Default Fallback) | Anthropic | 61.7 | $11.42 | 1.0h |
| 2 | Antigravity CLI - Gemini 4 Carbon (High) | 59.4 | $7.74 | 30m | |
| 3 | Claude Code - Claude Sonnet 5.5 (Max, Default Fallback) | Anthropic | 59.3 | $13.30 | 1.3h |
| 4 | Codex - GPT-6 Astra (Max) | OpenAI | 54.7 | $7.04 | 25m |
| 5 | Codex - GPT-6.1 Sol (Max) | OpenAI | 54.1 | $1.50 | 24m |
| 6 | Muse Code - Muse Wonder (Max) | Meta | 52.0 | $8.05 | 25m |
| 7 | Mistral Vibe - Mistral Large 4 (High) | Mistral | 48.3 | $1.26 | 48m |
| 8 | Grok Build - Grok 4.7 (Xhigh) | SpaceXAI | 46.6 | $9.48 | 36m |
| 9 | Muse Code - Muse Spark 1.3 (Max) | Meta | 45.7 | $3.91 | 23m |
| 10 | Muse Code - Muse Spark Open (Xhigh) | Meta | 26.3 | $3.54 | 26m |