Grok 4.6 released
xAI · Model release
Source confirmed
Data as of September 25, 2026 · How the score is built
Superseded.xAI has newer models in this line:Grok 4.7
SupersededProprietaryReasoning
Released Aug 12, 2026500K contextxAI Grok 4.6 model documentation
Grok 4.6
Decision readingGrok 4.6 scores 69.2 out of 100 and ranks #15 of 194. This profile shows 17 source-displayable benchmark rows; its strongest eligible category is Agentic at #8. API pricing is $2 input and $6 output per million tokens, with cached input at $0.5.
Released Aug 12, 2026 — see all recent releases
Grok 4.6 will be repriced, updated or retired. Get each notice with its source and date. Follow model changes
Share or export
Decision snapshot
Each value carries a field reference instead of floating alone. Markers compare this model with the current ranked and priced catalog; they are not absolute quality thresholds.
Capability
69.2/100
field median 50.2#15 of 194 ranked models
Public
#15of 194
Verified #11 of 74
Price
$2input / $6 output
input median $0.97cached $0.50 · blended $4
Speed
58tok/s
field median 91 tok/sFirst token 42.75 s
Context
500Ktokens
field median 256,000Maximum output length is tracked separately
Strongest published evidence
Agentic ranks #8. Particularly useful for coding agents, browser research, and computer-use workflows.
Validate before choosing
17 published rows leave some tracked benchmark slots empty. Reasoning is its lowest eligible category at #16.
Source-linked · 17 displayable benchmark rows
Follow model changesCategory score record
Scores and ranks appear only where published evidence can be displayed. The table keeps the score, weight, cohort, and evidence state together.
| Category | Score | Rank | Percentile | Weight | Benchmarks | Evidence |
|---|---|---|---|---|---|---|
| AgenticRank #8 of 105Percentile 93rdWeight 22%4 benchmarksVerified | 67.9 | 4 benchmarks | Verified | |||
| CodingRank #14 of 135Percentile 90thWeight 20%8 benchmarksVerified | 62.0 | 8 benchmarks | Verified | |||
| ReasoningRank #16 of 19Percentile 17thWeight 17%3 benchmarksVerified | 57.6 | 3 benchmarks | Verified | |||
| MultimodalWeight 12%0 benchmarksNot measured | Not measured | 0 benchmarks | Not measured | |||
| KnowledgeRank #13 of 158Percentile 92ndWeight 12%2 benchmarksVerified | 69.0 | 2 benchmarks | Verified | |||
| MultilingualWeight 7%0 benchmarksNot measured | Not measured | 0 benchmarks | Not measured | |||
| Inst. FollowingWeight 5%0 benchmarksNot measured | Not measured | 0 benchmarks | Not measured | |||
| MathWeight 5%0 benchmarksNot measured | Not measured | 0 benchmarks | Not measured |
17 of 486 tracked benchmark slots have displayable evidence · bars run 0–100
Coverage detailsHow much of this is verified
Coverage is split by category so a strong number never hides a thin evidence base. Verified means the row is tied to a published source; provisional rows remain visible but separate.
- Agentic4/4 verified
- Coding8/8 verified
- Reasoning3/3 verified
- MultimodalNot measured
- Knowledge2/2 verified
- MultilingualNot measured
- Inst. FollowingNot measured
- MathNot measured
Verified sourceProvisionalNot measured
Capability shape
Each axis shows percentile within that category’s eligible cohort. The comparison outline is the median of the six nearest public-score peers; a collapsed vertex means the category is not rank-eligible.
Grok 4.6 category percentile values
- Agentic93rd percentile
- Coding90th percentile
- Reasoning17th percentile
- MultimodalNot eligible
- Knowledge92nd percentile
- MultilingualNot eligible
- Instruction followingNot eligible
- MathNot eligible
The dashed outline is median of 6 nearest peers.
Eligible category ranks
- Agentic#8/105
- Coding#14/135
- Reasoning#16/19
- MultimodalNot ranked
- Knowledge#13/158
- MultilingualNot ranked
- Inst. FollowingNot ranked
- MathNot ranked
Top decileTop quartileMid-fieldNot eligible
Benchmark ledger
Coding opens by default. The marker compares each value with the best source-verified result in the catalog; provisional leaders do not set the reference. Expand the remaining categories for every published row.
Coding8 rows
| Benchmark | Score | Versus best verified row | Gap | Weight | Evidence |
|---|---|---|---|---|---|
| FrontierSWE v2 | Score25.3% | Versus best verified row Best verified: GPT-6 Astra · 65.5% | Gap40.2 behind | Weight8% ref. weight | Benchmark exact |
| LiveCodeBench (Vals)LiveCodeBench, Vals AI run | Score88.2% | Versus best verified row Best verified: Claude Fable 5.1 · 90.5% | Gap2.3 behind | Weight8% ref. weight | |
| VulcanBench v3 | Score87.0% | Versus best verified row Best verified: Grok 4.5 · 89.9% | Gap2.9 behind | Weight3% ref. weight | Benchmark exact |
| DeepSWE | Score65.9% | Versus best verified row Best verified: Muse Spark 1.3 · 75.4% | Gap9.5 behind | WeightDisplay only | Provider exact |
| CursorBench 3.2 | Score70.8% | Versus best verified row Best verified: Claude Fable 5.1 · 73.4% | Gap2.6 behind | WeightDisplay only | Benchmark exact |
| Bug Hunt Bench | Score27 fixes | Versus best verified row Best verified: Claude Fable 5.1 · 43 fixes | Gap16 behind | WeightDisplay only | Benchmark exact |
| FrontierCode 1.1 Extended | Score61.3% | Versus best verified row Best verified: GPT-6 Astra · 64.5% | Gap3.2 behind | WeightDisplay only | Provider exact |
| SWE-bench (Vals)SWE-bench, Vals AI run | Score95.6% | Versus best verified row Best verified: Claude Opus 5 · 97.0% | Gap1.4 behind | WeightDisplay only | Verified |
Agentic4 rows
| Benchmark | Score | Versus best verified row | Gap | Weight | Evidence |
|---|---|---|---|---|---|
| Terminal-Bench 2.1 (Vals)Terminal-Bench 2.1, Vals AI run | Score78.3% | Versus best verified row Best verified: GPT-6 Astra · 87.3% | Gap9 behind | Weight3% ref. weight | |
| Terminal-Bench 3.0 | Score26.5% | Versus best verified row Best verified: Claude Opus 5 · 42.7% | Gap16.2 behind | WeightDisplay only | Benchmark exact |
| APEX-Agents | Score57.5% | Versus best verified row Best verified: Grok 4.6 · 57.5% | GapBest verified | WeightDisplay only | Provider exact |
| ApprenticeBenchApprenticeBench: end-to-end computer use, continual learning, and long-horizon agency on a real accounts-payable job | Score13% | Versus best verified row Best verified: Claude Fable 5.1 · 72% | Gap59 behind | WeightDisplay only | Benchmark exact |
Reasoning3 rows
| Benchmark | Score | Versus best verified row | Gap | Weight | Evidence |
|---|---|---|---|---|---|
| ARC-AGI-2Abstraction and Reasoning Corpus for AGI v2 | Score67.1% | Versus best verified row Best verified: GPT-6 Astra · 95% | Gap27.9 behind | WeightWeighted 25% | Benchmark exact |
| ARC-AGI-3Abstraction and Reasoning Corpus for AGI v3 | Score2.1% | Versus best verified row Best verified: GPT-6 Astra · 62.7% | Gap60.6 behind | WeightWeighted 15% | Benchmark exact |
| ARC-AGI-1ARC-AGI-1 Semi-Private Evaluation | Score87.00% | Versus best verified row Best verified: GPT-6 Astra · 98.50% | Gap11.5 behind | WeightDisplay only | Benchmark exact |
Knowledge2 rows
| Benchmark | Score | Versus best verified row | Gap | Weight | Evidence |
|---|---|---|---|---|---|
| MMLU-Pro (Vals)MMLU-Pro, Vals AI run | Score89.4% | Versus best verified row Best verified: Claude Fable 5.1 · 92.4% | Gap3 behind | Weight6% ref. weight | Verified |
| GPQA Diamond (Vals)GPQA Diamond, Vals AI run | Score94.7% | Versus best verified row Best verified: Gemini 3.1 Pro · 95.5% | Gap0.8 behind | Weight2% ref. weight |
What it costs to get this score
Published API price against the public score. The x-axis uses a log scale; the dashed path marks models that are not beaten by a cheaper, higher-scoring option. Price uses average of published input and output rates.
Current modelExplore all models
Grok 4.6 · 69.2 score · $4 blended per million tokens
The chart opens on the current model. Scroll horizontally to inspect the full price axis.
Horizontal: blended price per million tokens, log scale · Vertical: public score
Lineage
The sequence follows explicit supersedes links. Each score is estimated for that model; a relative can inform a sparse estimate but never sets a floor, so a newer release can score below an earlier one. Scores and prices remain blank when the corresponding public row or first-party rate is unavailable.
Radar
Full release historyGrok 4.6 release history
Radar confirmed these at the source. Use Grok 4.6 in your work? Explore Radar to follow supported changes and choose your alerts.
Radar
Spec sheet
Each documented value carries its source. Missing fields stay visible as not sourced or not published, rather than disappearing from the page.
- API model ID
- grok-4.6xAI Grok 4.6 release notes
- Context window
- 500KxAI Grok 4.6 release notes
- Maximum output
- Not sourced yet
- Knowledge cutoff
- Not sourced yet
- Input modalities
- Not sourced yet
- Output modalities
- Not sourced yet
- Parameters
- Not disclosed by the provider
- Availability
- Available through the xAI API, Grok Build, Cursor, and supported model gateways. The API accepts text and image input and returns text.
- Cloud regions
- Not tracked yet
- Lifecycle
- Superseded
- API capabilities
- Tool calling, structured outputs, and batch support are not tracked yet
- Prompt caching
- Published at $0.50 per million cached input tokensxAI Grok 4.6 release notes
- Self-host
- Weights are not published
- Rate limits
- Not tracked yet
How to read this profile
The visual layer above carries the decisions. These notes preserve the model, ranking, coverage, and family context behind the numbers.
Grok 4.6 ranks #15 of 194 on the public leaderboard with a score of 69.19/100. Its source-verified position is #11 of 74.
Grok 4.6 is a proprietary model with a 500K context window. It uses an explicit reasoning mode, which can improve complex problem solving while adding latency and token use.
Available through the xAI API, Grok Build, Cursor, and supported model gateways. The API accepts text and image input and returns text.
Official exact-value snapshot from xAI's August 12, 2026 Grok 4.6 documentation and launch post, Cursor's live CursorBench 3.2 leaderboard, the official Terminal-Bench 3.0 leaderboard, and Bug Hunt Bench's blind-graded production-repository evaluation. Grok 4.6 is the successor to Grok 4.5, with a 500K-token context window and low, medium, high, or xhigh reasoning effort. We store xAI's published High results for DeepSWE 1.1, FrontierCode 1.1 Extended, and APEX-Agents; Cursor's Extra High row; Terminal-Bench's Grok Build row; and Bug Hunt Bench's verified xhigh ceiling. These results stay display-only because each depends on an evaluation or agent harness. The composite intelligence, GDPVal v2, APEX-SWE, professional-work, and legal-agent rows stay outside the core schema rather than being collapsed into non-equivalent metrics.
Its explicit predecessor is Grok 4.5. 17 of 486 tracked benchmark slots currently have displayable evidence. Missing categories stay blank.
Its strongest eligible category is Agentic at #8, while its lowest eligible position is Reasoning at #16. particularly useful for coding agents, browser research, and computer-use workflows.
Last updated September 25, 2026. Runtime fields remain blank until a sourced snapshot exists.
Questions
How does Grok 4.6 perform overall in AI benchmarks?
Grok 4.6 ranks #15 out of 194 models on the public BenchAlign leaderboard, with a score of 69.19/100. Its evidence status is Supported, and this profile shows 17 source-displayable benchmark rows. The label describes evidence depth, not a provider quality claim; inspect category rows before choosing a workload.
Is Grok 4.6 good for knowledge and understanding?
Grok 4.6 ranks #13 out of 158 eligible models for knowledge and understanding, with a public category score of 69/100. Higher-ranked alternatives are available for workloads where this category decides the choice. Check the underlying rows before treating the aggregate as a workload guarantee.
Is Grok 4.6 good for coding and programming?
Grok 4.6 ranks #14 out of 135 eligible models for coding and programming, with a public category score of 62/100. Higher-ranked alternatives are available for workloads where this category decides the choice. Check the underlying rows before treating the aggregate as a workload guarantee.
Is Grok 4.6 good for reasoning and logic?
Grok 4.6 ranks #16 out of 19 eligible models for reasoning and logic, with a public category score of 57.6/100. Higher-ranked alternatives are available for workloads where this category decides the choice. Check the underlying rows before treating the aggregate as a workload guarantee.
Is Grok 4.6 good for agentic tool use and computer tasks?
Grok 4.6 ranks #8 out of 105 eligible models for agentic tool use and computer tasks, with a public category score of 67.9/100. That places it in the current top ten for this category. Check the underlying rows before treating the aggregate as a workload guarantee.
Does Grok 4.6 have full benchmark coverage on BenchLM?
No. Grok 4.6 currently has 37 source-displayable rows across 486 tracked benchmark slots. The profile exposes published, non-generated evidence and leaves missing categories blank until an exact evaluation is available. Coverage describes how much was measured; it is not a penalty added to an individual benchmark result.
What is the context window size of Grok 4.6?
Grok 4.6 has a documented context window of 500K. That figure is the maximum combined prompt and retained-conversation space reported for this exact model; it is not the maximum output length. The profile keeps output limits separate because providers often publish those limits independently.