Model Details
Grok 4.6
Quick answer: Grok 4.6 is xAI's August 2026 flagship model, focused on agentic coding and long-horizon task execution. It reports DeepSWE 65.9%, APEX-Agents 57.5%, AA Briefcase 1577 Elo, GDPVal-AA 1753 Elo, and an AA Intelligence Index of 61 — priced at $2 input / $6 output per 1M tokens, with a faster variant available at 2x that price.
What Grok 4.6 Is
Grok 4.6 is xAI's incremental flagship release following Grok 4.5, announced August 12, 2026. The release is positioned around agentic coding and tool-use reliability rather than raw knowledge benchmarks — xAI's announcement leads with scores on agent-native evals like DeepSWE, APEX-Agents, CursorBench, and FrontierCode rather than static knowledge tests.
The model is available through Cursor, xAI's Grok Build product, the xAI API, OpenRouter, Vercel AI Gateway, and Cloudflare Workers AI at launch — a distribution footprint aimed squarely at developer coding tools rather than consumer chat.
Specifications
| Field | Value |
|---|---|
| Organization | xAI |
| License | Proprietary (API only) |
| Release date | August 12, 2026 |
| Modality | Multimodal |
| Knowledge cutoff | February 2026 |
Public Benchmark Scores
| Benchmark | Score | Source | Date |
|---|---|---|---|
| DeepSWE | 65.9% | xAI Grok 4.6 announcement | 2026-08 |
| APEX-Agents | 57.5% | xAI Grok 4.6 announcement | 2026-08 |
| AA Briefcase | 1577 Elo | xAI Grok 4.6 announcement | 2026-08 |
| GDPVal-AA v2 | 1753 Elo | xAI Grok 4.6 announcement | 2026-08 |
| CursorBench | 69.9% | xAI Grok 4.6 announcement | 2026-08 |
| FrontierCode (Extended) | 61.3% | xAI Grok 4.6 announcement | 2026-08 |
| APEX-SWE | 56.4% | xAI Grok 4.6 announcement | 2026-08 |
| Terminal-Bench 3.0 | 26% | xAI Grok 4.6 announcement | 2026-08 |
xAI also reports an AA Intelligence Index score of 61 — Artificial Analysis's composite of nine evaluations. Benchgen does not maintain a standalone leaderboard page for this composite index; the score is cited here for completeness.
Scores as reported in xAI's own Grok 4.6 announcement (x.ai/news/grok-4-6, August 2026); not independently verified by Benchgen.
Pricing
| Variant | Input (per 1M tokens) | Output (per 1M tokens) |
|---|---|---|
| Grok 4.6 | $2.00 | $6.00 |
| Grok 4.6 Fast | $4.00 | $12.00 |
Available via the xAI API, OpenRouter, Vercel AI Gateway, and Cloudflare Workers AI.
Grok 4.6 vs Alternatives
| Model | DeepSWE | AA Briefcase | Price (in/out per 1M) |
|---|---|---|---|
| Grok 4.6 | 65.9% | 1577 Elo | $2 / $6 |
| Grok 4.5 | — | — | $2 / $6 |
Use Grok 4.6 via API
from openai import OpenAI # xAI API is OpenAI-compatible
client = OpenAI(api_key="YOUR_API_KEY", base_url="https://api.x.ai/v1")
response = client.chat.completions.create(
model="grok-4-6",
messages=[{"role": "user", "content": "Refactor this function for readability..."}],
)
print(response.choices[0].message.content)Related Models & Benchmarks
- Compare: Grok 4.5 · Grok 4
- Benchmarks: DeepSWE · APEX-Agents · AA Briefcase · GDPVal-AA v2
Frequently Asked Questions
What is Grok 4.6?
Grok 4.6 is xAI's August 2026 flagship model, focused on agentic coding and long-horizon task execution, succeeding Grok 4.5.How much does Grok 4.6 cost?
$2 per 1M input tokens and $6 per 1M output tokens; a faster variant is priced at $4 / $12 per 1M tokens.Is Grok 4.6 open source?
No — Grok 4.6 is proprietary and available only via API.What is Grok 4.6's knowledge cutoff?
February 2026.Specs and scores sourced from xAI's official Grok 4.6 announcement. Last updated 2026-08-12.
Benchmark Leaderboards
This model isn’t on any benchmark leaderboard yet.