Models · Best vision
Grok 4
by xAI
Grok model for agentic tool use, reasoning, coding, and live assistance.
Our verdict
We have not reviewed Grok 4 yet
No one at Enterprise DNA has checked this listing against its sources. The description comes from public listings gathered automatically. Any fact below shows where it came from and when it was last checked.
From the listing
About Grok 4
Grok model for agentic tool use, reasoning, coding, and live assistance.
Notes
Grok 4 is listed in the EDNA Models directory, sourced from models.dev.
Key facts
What we have checked
- Time to first token (ms)
- 11,220
- Output speed (tokens a second)
- 73.7
- Gpqa diamond
- 87% (for 9 July 2025)
- Terminal bench
- 27.2% (for 9 July 2025)
- Arc agi 2
- 16% (for 9 July 2025)
- Aider polyglot
- 79.6% (for 9 July 2025)
- Arc agi
- 66.7% (for 9 July 2025)
- Balrog
- 43.6% (for 9 July 2025)
- Cybench
- 43% (for 9 July 2025)
- Deepresearch bench
- 47.3% (for 9 July 2025)
- Fiction livebench
- 94.4% (for 9 July 2025)
- Frontiermath 2025 02 28 private
- 19.7% (for 9 July 2025)
Each value shows when we last confirmed it and where it came from. Values we have not confirmed are left out, never guessed.
API access
How to call it
Call via other with model id `xai/grok-4`.
- other ·
xai/grok-4
Scores
How Grok 4 scores
Each score comes from the named source and shows when we last confirmed it. A flag means the score needs care.
| Score | Value | Source | Checked |
|---|---|---|---|
| Aider polyglot (best-across-versions) | 79.6percentAs of 9 July 2025 | Epoch AI (CC BY 4.0) | |
| Arc agi (best-across-versions) | 66.7percentAs of 9 July 2025 | Epoch AI (CC BY 4.0) | |
| Arc agi 2 (best-across-versions) | 16percentAs of 9 July 2025 | Epoch AI (CC BY 4.0) | |
| Balrog (best-across-versions) | 43.6percentAs of 9 July 2025 | Epoch AI (CC BY 4.0) | |
| Cybench (best-across-versions) | 43percentAs of 9 July 2025 | Epoch AI (CC BY 4.0) | |
| Deepresearch bench (best-across-versions) | 47.3percentAs of 9 July 2025 | Epoch AI (CC BY 4.0) | |
| Fiction livebench (best-across-versions) | 94.4percentAs of 9 July 2025 | Epoch AI (CC BY 4.0) | |
| Frontiermath 2025 02 28 private (best-across-versions) | 19.7percentAs of 9 July 2025 | Epoch AI (CC BY 4.0) | |
| Frontiermath tier 4 2025 07 01 private (best-across-versions) | 2.1percentAs of 9 July 2025 | Epoch AI (CC BY 4.0) | |
| Gdpval (best-across-versions) | 21.1percentAs of 9 July 2025 | Epoch AI (CC BY 4.0) | |
| Geobench (best-across-versions) | 45percentAs of 9 July 2025 | Epoch AI (CC BY 4.0) | |
| Gpqa diamond (best-across-versions) | 87percentAs of 9 July 2025 | Epoch AI (CC BY 4.0) |
Show 6 more scores
| Score | Value | Source | Checked |
|---|---|---|---|
| Lech mazur writing (best-across-versions) | 76.9percentAs of 9 July 2025 | Epoch AI (CC BY 4.0) | |
| Metr time horizons (best-across-versions) | 66.6percentAs of 9 July 2025 | Epoch AI (CC BY 4.0) | |
| Otis mock aime 2024 2025 (best-across-versions)Saturated | 84percentAs of 9 July 2025 | Epoch AI (CC BY 4.0) | |
| Simplebench (best-across-versions) | 60.5percentAs of 9 July 2025 | Epoch AI (CC BY 4.0) | |
| Terminal bench (best-across-versions) | 27.2percentAs of 9 July 2025 | Epoch AI (CC BY 4.0) | |
| Weirdml (best-across-versions) | 45.7percentAs of 9 July 2025 | Epoch AI (CC BY 4.0) |
- Saturated
- Top models score close to the maximum, so this test no longer separates them.
Compare
Compare Grok 4 side by side
Sources
Where this comes from
- Artificial AnalysisThird party
- Epoch AI (CC BY 4.0)Third party
- Epoch AI (CC BY 4.0)Third party
- Epoch AI (CC BY 4.0)Third party
- Epoch AI (CC BY 4.0)Third party
- Epoch AI (CC BY 4.0)Third party
- Epoch AI (CC BY 4.0)Third party
- Epoch AI (CC BY 4.0)Third party
- Epoch AI (CC BY 4.0)Third party
- Epoch AI (CC BY 4.0)Third party
- Epoch AI (CC BY 4.0)Third party
A source shows what was published when we checked it. It does not certify security, performance or a working installation. Read our method.
Changelog
What changed
No changes recorded yet. Releases, price changes and deprecations appear here, with their dates, once we track them for this record.
MODEL FIELD GUIDE / SOURCE EVIDENCE
What you can do with Grok 4.
Listed specifications; no EDNA task evaluation has been run. Sources checked 2026-11-07. Hosting routes and app features can differ.
xai/grok-4- Tool calling
- Listed support
- Structured output
- Not established
- Reasoning controls
- Listed support
- Availability
- available
- Price basis
- token
- Licence verification
- Check the developer licence; the model name is insufficient
- Context length
- 256K tokens
- Knowledge cutoff
- Not established
- Developer release date
- Not established
- First seen in the catalogue
- Not established
- Listed retirement date
- Not established
Inspect the sources and their scope
xai/grok-4: listed by 2 providers on models.dev (llmgateway-providers, poe); a fact is kept only where the listings agree. Hosting routes can differ.
Checked 2026-11-07Source for reasoning controls, tool calling, context length, input, output.
Checked 2026-11-07FROM SPECIFICATIONS TO YOUR WORK
Is this the right model for your job?
Start with a realistic sample. These workflows match the model’s listed capabilities, not a verified result on your task.
What would it cost?
Pick a common job and how many times a month you run it. Token sizes are our assumption for a typical run, shown under the result.
Enter a whole number of runs from 1 to 100,000,000.
Working out the cost...
Token prices only. Per-request, image, tool, storage and subscription fees are not included. One token is about three quarters of an English word. Real counts vary by model and language. How tokens are counted ↗
Check the source
Price date: 13 Jul 2026, from the models.dev listing. Prices are USD API token rates. Your app subscription, uploads, browsing and tools are separate.
Model source & availability ↗Understand the evidence
Listed capabilities and third-party benchmark scores do not establish Enterprise DNA task-test results. Validate accuracy and total cost on your own sample.
Intelligence Index: 33.3 (2026-07-21) ↗Get the free Developer’s Field Guide
A 27-page field guide to the AI coding workflow with Claude. Claude Code, MCP servers, the prompt patterns that work, and what to delegate. Free.
Enter your work email. We send it straight over, plus a few short notes worth knowing. Unsubscribe any time.