xAIReasoning modelGrok 4
Best configuration ranks #144 of 610 on the BenchLeader Index at 57.2 ±3.0. Last measured 2 Sept 2026. Released 9 Jul 2025.
- Blended price
- $1.56/M
- $1.25 in · $2.50 out
- Output speed
- –
- First answer
- –
- Context
- 256k
- Overall index57
- Reasoning58
- Coding60
- Agents & tools55
- Maths53
- Knowledge58
- Instruction following54
- Human preference60
- Multimodal54
- Long context69
- Composite57
Reasoning-effort configurations
The same model behaves differently depending on how much it is allowed to think. Each row is one setting, scored only on the benchmarks that were run at that setting. “Default” means the publisher did not say which setting was used.
| Effort | Index | Rank | Speed | First answer | Chat reply cost | Categories |
|---|---|---|---|---|---|---|
| high | – | – | – | – | $0.0057 | Agents & tools 34 |
| thinking | – | – | – | – | $0.0057 | Reasoning 45 |
| defaultbest | 57.2 | #144 | – | – | $0.0013 | Agents & tools 55 · Coding 60 · Composite 57 · Human preference 60 · Instruction following 54 · Knowledge 58 · Long context 69 · Maths 53 · Multimodal 54 · Reasoning 58 |
Benchmark results
One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.
Reasoning
| Benchmark | high | thinking | default | Source |
|---|---|---|---|---|
| GPQA Diamond | – | – | 87.0%#64 | Epoch AI Benchmarking Hub |
| SimpleBench | – | – | 60.5%#30 | SimpleBench |
| LMArena Hard Prompts | – | – | 1420#128 | LMArena |
| GPQA Diamond (AA)not in index | – | – | 87.7%#94 | Artificial Analysis |
| Humanity's Last Exam (AA)not in index | – | – | 26.7%#129 | Artificial Analysis |
| GPQA Diamond (Vals)not in index | – | – | 88.1%#34 | Vals AI |
| Kagi LLM Benchmark | – | – | 73.6%#17 | Kagi LLM Benchmark |
| ARC-AGI-1 | – | 66.7%#100 | 79.6%#82 | ARC Prize |
| ARC-AGI-2 | – | 16.0%#98 | 29.4%#89 | ARC Prize |
Coding
| Benchmark | high | thinking | default | Source |
|---|---|---|---|---|
| WeirdML | – | – | 45.7%#80 | WeirdML |
| LMArena Coding | – | – | 1435#144 | LMArena |
| LiveCodeBench | – | – | 83.3%#48 | Vals AI |
| IOI | – | – | 26.2%#17 | Vals AI |
| SWE-bench (Vals)not in index | – | – | 57.8%#75 | Vals AI |
| Aider Polyglot | – | – | 79.6%#5 | Aider polyglot leaderboard |
Agents & tools
| Benchmark | high | thinking | default | Source |
|---|---|---|---|---|
| Terminal-Bench | – | – | 27.2%#47 | Terminal-Bench |
| GDPval | 21.1%#10 | – | – | OpenAI |
| Cybench | – | – | 43.0%#4 | Cybench |
| APEX-Agents | – | – | 15.2%#46 | Mercor |
| Terminal-Bench Hard | – | – | 37.9%#61 | Artificial Analysis |
| τ²-Bench Telecom (AA)not in index | – | – | 74.8%#134 | Artificial Analysis |
| BFCL Overall | – | – | 63.0%#8 | Berkeley Function Calling Leaderboard |
Maths
| Benchmark | high | thinking | default | Source |
|---|---|---|---|---|
| OTIS Mock AIME | – | – | 84.0%#89 | Epoch AI Benchmarking Hub |
| AIME (Vals) | – | – | 90.6%#27 | Vals AI |
| MGSM | – | – | 90.9%#31 | Vals AI |
| IMO 2025 | – | – | 21.4%#3 | MathArena |
| MathArena Apex | – | – | 2.1%#29 | MathArena |
Knowledge
| Benchmark | high | thinking | default | Source |
|---|---|---|---|---|
| AA-Omniscience | – | – | 2.1#82 | Artificial Analysis |
| MMLU-Pro | – | – | 85.3%#57 | Vals AI |
| LegalBench | – | – | 83.2%#51 | Vals AI |
| CorpFin | – | – | 66.0%#23 | Vals AI |
| TaxEval | – | – | 65.1%#112 | Vals AI |
| MedQA | – | – | 92.5%#31 | Vals AI |
Instruction following
| Benchmark | high | thinking | default | Source |
|---|---|---|---|---|
| IFBench | – | – | 53.7%#153 | Artificial Analysis |
Human preference
| Benchmark | high | thinking | default | Source |
|---|---|---|---|---|
| LMArena Text | – | – | 1411#121 | LMArena |
Multimodal
| Benchmark | high | thinking | default | Source |
|---|---|---|---|---|
| LMArena Vision | – | – | 1210#69 | LMArena |
| MMMU-Pro | – | – | 68.8%#128 | Artificial Analysis |
Long context
| Benchmark | high | thinking | default | Source |
|---|---|---|---|---|
| Fiction.LiveBench 120k | – | – | 96.9%#2 | Fiction.live |
| AA-LCR | – | – | 68.0%#180 | Artificial Analysis |
Composite
| Benchmark | high | thinking | default | Source |
|---|---|---|---|---|
| Epoch Capabilities Indexnot in index | – | – | 146.4#62 | Epoch AI Benchmarking Hub |
| AA Intelligence Index | – | – | 22.5#143 | Artificial Analysis |
Where it wins
Benchmarks where this configuration ranks in the top five of every configuration measured.
What a task costs
Estimates from list price, output speed and time to first answer for the best configuration. “With caching” assumes three-quarters of the input is served from the prompt cache. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0013 | – | – |
| Summarise a 30-page report | 12,000 / 600 | $0.017 | – | – |
| Code edit | 6,000 / 1,500 | $0.011 | – | – |
| Agentic coding session | 60,000 / 4,000 | $0.085 | – | – |
| Structured extraction | 2,000 / 200 | $0.0030 | – | – |
See also
Data as of 9 Sept 2026. Compare these configurations.