SpaceXAIGrok 3
Grok 3 is a SpaceXAI proprietary model, released 17 Feb 2025. Its best configuration ranks #340 of 785 on the BenchLeader Index at 51.5 ±3.4, in the upper half. The ± is the point: 295 other configurations score within that range, so they and this one cannot be told apart on quality alone — price and speed are what separate them. It scores highest in multimodal (65) and lowest in composite (43). At $6.00 per million tokens blended it is among the most expensive ranked models. Output speed of 91 tokens per second puts it faster than most, with a first answer in 1.6 s. It has been measured at 2 reasoning-effort settings; this summary describes the best-scoring one, and the tabs above switch between them. Last measured 22 Oct 2026.
- Blended price
- $6.00/M
- $3.00 in · $15.00 out
- Output speed
- 91 tok/s
- measured by Artificial Analysis
- First answer
- 1.61 s
- first token 1.61 s
- Context
- 1M
- Full answer
- 7.12 s
- median, reasoning included
- Cost per run
- –
- Released
- 17 Feb 2025
- Overall index52
- Reasoning43
- Coding47
- Agents & tools44
- Maths55
- Knowledge51
- Instruction following57
- Human preference59
- Multimodal65
- Long context52
- Composite43
Versions
SpaceXAI has shipped 11 models under this name. Each is ranked on its own results; a newer version often has fewer results so far, which holds its index nearer the average until more arrive.
| Model | Released | Index | Rank |
|---|---|---|---|
| Grok 4.7high | 21 Sept 2026 | 63.9 | #59 |
| Grok 4.6medium | 11 Aug 2026 | 64.2 | #56 |
| Grok 4.5high | 8 Jul 2026 | 59.3 | #134 |
| Grok 4.3medium | 17 Apr 2026 | 58.7 | #147 |
| Grok 4.2high | 10 Mar 2026 | – | – |
| Grok 4.20thinking | 17 Feb 2026 | 58.6 | #148 |
| Grok 4.1 | 17 Nov 2025 | 57.2 | #184 |
| Grok 4 | 9 Jul 2025 | 58.0 | #161 |
| Grok 3this page | 17 Feb 2025 | 51.5 | #340 |
| Grok 2 | 12 Dec 2024 | 41.9 | #612 |
| Grok 4.20 1 | – | – | – |
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. “Not stated” collects results from publishers that did not say which setting they used; for a reasoning model that is usually its thinking mode, but we do not assume it. Pick a setting here or at the top of the page to see its price, speed and category scores.
| Effort | Index | Rank | Speed | First answer | Chat reply cost | Categories |
|---|---|---|---|---|---|---|
| thinking | – | – | 91 tok/s | 1.61 s | $0.0057 | Composite 41 |
| not statedbest | 51.5 | #340 | 91 tok/s | 1.61 s | $0.0057 | Agents & tools 44 · Coding 47 · Composite 43 · Human preference 59 · Instruction following 57 · Knowledge 51 · Long context 52 · Maths 55 · Multimodal 65 · Reasoning 43 |
How its index has moved
8 Sept 2026 to 18 Oct 2026The index is recomputed from scratch every day, so a line moves when a new benchmark result lands, when a publisher revises a score, or when the models it is normalised against change. Early movement usually means the score is still settling.
Benchmark results
See all LLM benchmarks. 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 | not stated | Source | Trend |
|---|---|---|---|
| GPQA Diamond | 75.8%#144 | Epoch AI Benchmarking Hub | |
| SimpleBench | 36.1%#63 | SimpleBench | |
| Arena Hard Prompts | 1426#142 | Arena (LMArena) | |
| GPQA Diamond (AA)not in index | 69.3%#306 | Artificial Analysis | |
| Humanity's Last Exam (AA)not in index | 4.1%#538 | Artificial Analysis | |
| GPQA Diamond (Vals)not in index | 74.2%#81 | Vals AI | |
| Kagi LLM Benchmark | 61.3%#47 | Kagi LLM Benchmark | |
| GPQA Diamond (HELM)not in index | 65.0%#20 | HELM Capabilities | |
| ARC-AGI-1 | 5.5%#243 | ARC Prize | |
| ARC-AGI-2 | 0.0%#242 | ARC Prize | |
| CritPt | 0.0%#315 | Artificial Analysis |
Coding
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| WeirdML | 37.2%#119 | WeirdML | |
| Arena Coding | 1442#154 | Arena (LMArena) | |
| LiveCodeBench | 52.9%#109 | Vals AI | |
| Aider Polyglot | 53.3%#22 | Aider polyglot leaderboard |
Agents & tools
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Terminal-Bench Hard | 11.4%#215 | Artificial Analysis | |
| τ²-Bench Telecom (AA)not in index | 48.8%#200 | Artificial Analysis | |
| BALROGnot in index | 29.5%#18 | BALROG | |
| BrowseCompnot in index | 1.3%#13 | OpenAI |
Maths
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| OTIS Mock AIME | 55.6%#174 | Epoch AI Benchmarking Hub | |
| MATH Level 5 | 88.8%#23 | Epoch AI Benchmarking Hub | |
| AIME (Vals) | 58.8%#56 | Vals AI | |
| MGSM | 91.3%#26 | Vals AI | |
| Omni-MATH (HELM) | 46.4%#23 | HELM Capabilities | |
| MATH 500not in index | 89.8%#27 | Vals AI | |
| Arena Maths | 1372#179 | Arena (LMArena) |
Knowledge
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| AA-Omniscience | -33.2#293 | Artificial Analysis | |
| MMLU-Pro | 80.0%#87 | Vals AI | |
| LegalBench | 82.6%#66 | Vals AI | |
| CorpFin | 59.7%#72 | Vals AI | |
| TaxEval | 75.9%#10 | Vals AI | |
| MedQA | 83.8%#60 | Vals AI | |
| MMLU-Pro (HELM)not in index | 78.8%#22 | HELM Capabilities | |
| AA-Omniscience: accuracynot in index | 28.9%#193 | Artificial Analysis | |
| AA-Omniscience: non-hallucinationnot in index | 12.7%#372 | Artificial Analysis |
Instruction following
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| IFBench | 46.9%#192 | Artificial Analysis | |
| IFEval (HELM) | 88.4%#12 | HELM Capabilities | |
| WildBench (HELM) | 84.9%#13 | HELM Capabilities | |
| Arena Instruction Followingnot in index | 1403#133 | Arena (LMArena) |
Human preference
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Arena Text | 1411#138 | Arena (LMArena) | |
| Arena Creative Writingnot in index | 1398#109 | Arena (LMArena) | |
| Arena Multi-turnnot in index | 1410#141 | Arena (LMArena) | |
| Arena Longer Queriesnot in index | 1428#124 | Arena (LMArena) | |
| Lech Mazur Creative Writingnot in index | 76.4%#23 | Lech Mazur |
Multimodal
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| MMMU (validation) | 78.0%#10 | MMMU |
Long context
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Fiction.LiveBench 120k | 58.3%#19 | Fiction.live | |
| AA-LCR | 58.0%#289 | Artificial Analysis |
Composite
| Benchmark | thinking | not stated | Source | Trend |
|---|---|---|---|---|
| Epoch Capabilities Indexnot in indexeffort not stated | – | 138.2#121 | Epoch AI Benchmarking Hub | |
| AA Intelligence Index v4.3.2 | 10.4#377 | 12.1#337 | Artificial Analysis | |
| HELM Capabilities meannot in indexeffort not stated | – | 72.7%#17 | HELM Capabilities |
Price history
The price above, per 1M tokens, every day it changed.
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 | $0.0057 | – | 4.9 s |
| Summarize a 30-page report | 12,000 / 600 | $0.045 | – | 8.2 s |
| Code edit | 6,000 / 1,500 | $0.041 | – | 18.1 s |
| Agentic coding session | 60,000 / 4,000 | $0.240 | – | 45.7 s |
| Structured extraction | 2,000 / 200 | $0.0090 | – | 3.8 s |
See also
Questions people ask
- What are Grok 3's benchmark scores?
- As of 24 Oct 2026, Grok 3 scores 51.5 ±3.4 on the BenchLeader Index, strongest in multimodal (65) and weakest in composite (43). That rests on 32 independent benchmark results, and every score on this page links to its publisher.
- How does Grok 3 rank?
- Grok 3 is #340 of 785 ranked configurations on 24 Oct 2026. 295 other configurations score within its margin, so the exact position matters less than the band. The leader is Claude Opus 5.5 (max effort) at 72.6.
- How much does Grok 3 cost per 1M tokens?
- As of 24 Oct 2026, Grok 3 costs $3.00 per 1M input tokens and $15.00 per 1M output tokens, $6.00 blended (three input tokens per output token). The list price last changed on 13 Oct 2026.
- When was Grok 3 released?
- SpaceXAI released Grok 3 on 17 Feb 2025. BenchLeader has results for it at 2 reasoning-effort settings.
Data as of 24 Oct 2026. Compare these configurations.
Cite as: BenchLeader, “Grok 3: benchmarks, pricing, speed and rank”, data as of 24 Oct 2026, https://www.benchleader.com/models/grok-3.