xAI released Grok 4.7 on September 21, 2026, its newest model for coding and knowledge work. The company says the model works longer on difficult tasks, checks its own work more carefully, and carries its best-calibrated safeguards to date — served at the same list price and speed as Grok 4.6, and positioned as highly competitive on price-performance.
Grok 4.7 is available today in Cursor and Grok Build, through the Grok API, third-party coding harnesses, and model routers and cloud platforms. Pricing starts at $2 per million input and $6 per million output tokens. A fast variant with twice the output speed at twice the price is also offered.
Confirmed
- New, larger base model than Grok 4.6, trained with a longer reinforcement-learning run on a harder mix of multi-hour tasks; better at verifying its own work, managing longer context, and natively understanding the Grok Bot harness for conversational and knowledge tasks.
- xAI's launch table (reproduced below) puts Grok 4.7 xHigh at 46.3% on CursorBench 4.0 versus 40.4% for Grok 4.6 High, with matched $2/$6 list pricing — and lower list prices than the GPT-5.6 Sol Max and Fable 5.1 Max rows in the same table.
- Improved document and presentation creation, with gains on the GDPval and AA Briefcase professional-work benchmarks versus Grok 4.6 (vendor scores).
- New safeguard stack: xAI says it is the strongest refusal and jailbreak resistance it has tested — tops LatchBio biosafety at 62.4% and allows only 3.3% of risky dual-use prompts on HackerBench v0.3 while rarely blocking legitimate security work; invite-only red-team access for select cybersecurity partners.
- Specs and access: 500,000-token context window; text and image input, text output; May 2026 knowledge cutoff; reasoning effort levels low, medium, high, xhigh (high default); function calling, structured outputs, web search, X search, code execution; Responses and Chat Completions APIs; Batch API unsupported.
- Tiered pricing in the model card: below 200,000 tokens — $2.00/M input, $0.50/M cached input, $6.00/M output; at or above 200,000 tokens — $4.00/M input, $1.00/M cached input, $12.00/M output, with the higher rates applying to all tokens once the threshold is crossed.
| Metric | Grok 4.7 xHigh | Grok 4.6 High | GPT-5.6 Sol Max | Fable 5.1 Max |
|---|---|---|---|---|
| Input $ / M tokens | $2 | $2 | $4 | $10 |
| Output $ / M tokens | $6 | $6 | $20 | $50 |
| CursorBench 4.0 | 46.3% | 40.4% | 41.7% | 51.8% |
| DeepSWE v1.1 | 71.0%* | 65.2% | 72.7% | 70.0% |
| EEBench | 64.0% | 53.0% | 39.4% | 56.4% |
| AA Briefcase v1.1 | 1,657 | 1,546 | 1,487 | 1,678 |
| Terminal-Bench 4.0 | 38.0% | 20.3% | 37.3% | 57.9% |
| Harvey Legal Agent | 19.6% | 15.8% | 2.5% | 6.7% |
| HealthBench Professional | 56.7% | 48.5% | 60.5% | 62.1% |
* high effort. Source: xAI Grok 4.7 launch — vendor figures; effort labels are not aligned across model families.
Unknown
- Parameter count not disclosed in the launch announcement or model pages.
- Independent replication of vendor-published benchmark scores (CursorBench, DeepSWE, Terminal-Bench, AA Briefcase, Harvey Legal, HealthBench Professional, EEBench, GDPval, HackerBench, LatchBio).
- Whether effort settings (xHigh vs High vs Max) are comparable across the models in xAI's table.
- Rollout scope and timing for third-party harnesses, model routers, and cloud platforms beyond Cursor and Grok Build.
- SLA, uptime guarantees, and enterprise support terms for the new model.
- Whether the fast variant is a distinct API model or a serving option only in Cursor and Grok Build.
Our take
xAI is positioning Grok 4.7 as a price-performance play for long-running agent workloads. The vendor table shows clear gains over Grok 4.6 at the listed effort settings, and the $2/$6 list price undercuts the Sol and Fable rows xAI chose for comparison. But the chart mixes effort labels across families, and the company's own HackerBench and LatchBio numbers are unaudited. Buyers should treat the launch scores as directional and run their own evaluation on representative tasks before committing production traffic.
Series: 1. xAI Ships Grok Imagine Image 2.0 With Precise Editing and Top Arena Ranking · 2. SpaceXAI Launches Grok Bot, Persistent AI Teammates With Their Own Cloud Computers · 3. Cursor Acquired by SpaceX, Grok 4.6 Released as First Joint Model · 4. X Open-Sourced the For You Algorithm: Not a Leak — a Transparency Push · 5. NVIDIA ships Vera CPU to Anthropic, OpenAI, Oracle, and SpaceXAI · 6. xAI launches Grok 4.7 for coding and knowledge work · xAI Grok
