Grok 4.7 API: coding, knowledge work, and pricing

TTAPI developers can now use the Grok 4.7 API through the platform's OpenAI-compatible chat completions route. xAI released Grok 4.7 on September 21, 2026 as its most capable model for coding and knowledge work, and TTAPI lists the exact model ID as grok-4.7.

The release positions Grok 4.7 for work that lasts longer than a single prompt: software engineering, terminal tasks, documents, presentations, electrical engineering, legal work, and clinical reasoning. Those are xAI product claims. They describe the model's intended scope rather than a TTAPI-run benchmark.

• Coding and multi-hour terminal work
• Documents, presentations, and professional analysis
• Longer context with more self-verification
• A new safeguard stack for dual-use domains

Grok 4.7 uses a larger base model than Grok 4.6 and a longer reinforcement-learning run on a harder task mix. xAI says the training placed more weight on problems that take hours to complete, with improvements in checking work and managing long context.

The model also natively understands xAI's Grok Bot harness. That matters for conversational tasks and general knowledge work, but a harness capability should not be confused with a plain API completion. Before adopting Grok 4.7, define whether your workload needs one response, a tool-using agent, or a long-running coding environment.

On xAI's published results, Grok 4.7 scores 46.3% on CursorBench 4.0 versus 40.4% for Grok 4.6. On DeepSWE v1.1, the reported high-effort result is 71.0%, up from 65.2%. Terminal-Bench 4.0 rises from 20.3% to 38.0%. These figures point toward better persistence across repository and terminal tasks, but they remain vendor-reported results under specific evaluation settings.

For a production coding workflow, test the tasks that actually fail today: multi-file changes, test repair, terminal recovery, long-context instruction retention, and final verification. Record latency, token use, failed tool calls, and review effort alongside whether the answer was technically correct.

• Use the same repository task and acceptance checks across models.
• Measure usable completion, not only first-response quality.
• Keep reasoning effort and tool access consistent during comparison.
• Review vendor benchmark claims separately from your own evaluation.

xAI also highlights documents, presentations, electrical engineering, legal work, and clinical reasoning. Grok 4.7 improves on Grok 4.6 in the published AA Briefcase and EEBench comparisons, and xAI describes performance comparable with other frontier models on professional tasks.

The practical question is whether the model can create an artifact that survives review. Test structure, factual traceability, spreadsheet or slide consistency, missing-information handling, and whether the output follows the requested format. A strong benchmark score does not remove the need for domain review in legal, clinical, financial, or engineering workflows.

xAI says Grok 4.7 uses an entirely new safeguard stack and is its strongest model so far for refusal quality and jailbreak resistance. The release reports a 62.4% result on LatchBio's biosafety benchmark and a 3.3% pass-through rate for risky dual-use prompts on HackerBench v0.3, while aiming to preserve utility for legitimate defensive work.

Treat those numbers as xAI's launch evidence. Teams handling cybersecurity or biological workflows still need scoped authorization, logging, human review, and task-level controls. Model-level safeguards are one layer in the system, not the complete operating policy.

At launch, xAI says standard Grok 4.7 is served at the same price and speed as Grok 4.6. A fast variant offers twice the output speed at twice the price. The standard model is available through the xAI API, Cursor, Grok Build, third-party coding harnesses, routers, and cloud platforms. Treat these as launch terms and confirm the current provider page before budgeting a production workload.

On TTAPI, send model grok-4.7 to POST /v1/chat/completions with the platform key in the Authorization header. The current catalog applies the xAI channel discount and links live billing from the pricing page. Start with one representative coding task, record response quality, latency, token usage, and review effort, then compare it with the model already serving that workload.

curl --request POST \
  --url 'https://api.ttapi.io/v1/chat/completions' \
  --header "Authorization: Bearer $TTAPI_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "model": "grok-4.7",
  "messages": [
    {
      "role": "user",
      "content": "Review this TypeScript API client. Identify the three highest-value reliability fixes and give one validation test for each."
    }
  ],
  "stream": false
}'

• Keep TTAPI_KEY in a server-side environment variable.
• Use the exact model ID grok-4.7 on /v1/chat/completions.
• Record response quality, latency, token usage, and request identifiers.
• Keep a rollback path to the model currently validated in production.