Grok 4.7
Language Available Full comparison ↗ xai/grok-4.7 · by xAI
· decoder-only-transformer
Pricing — 1 offering(s)
Input tokens
- $2.00 / 1M tokens (input)
Current
2026-09-21 → present
Standard-variant rate, confirmed via the launch announcement's own body text (raw HTML, not an AI-summarized…
Standard-variant rate, confirmed via the launch announcement's own body text (raw HTML, not an AI-summarized fetch). Same convention as grok-4-6-xai: xAI also serves a "fast" variant at 2x price ($4/1M) with 2x the output speed, noted here rather than tracked as a separate tier.
Output tokens
- $6.00 / 1M tokens (output)
Current
2026-09-21 → present
Standard-variant rate (see input tier's note for the fast-variant caveat: $12/1M at 2x output speed, not…
Standard-variant rate (see input tier's note for the fast-variant caveat: $12/1M at 2x output speed, not tracked as a separate tier).
Showing the active price and any recorded history. Full pricing history is available via the paid API — see API docs.
See how Grok 4.7 fits into a cost-aware routing setup
See how →Capability profile
How Grok 4.7 rates across core capability dimensions, with the task-level evidence behind each rating.
"Works longer on difficult tasks, checks its own work more carefully" per xAI's own launch framing — explicit emphasis on self-verification over raw speed, continuing Grok 4.6's agentic-reasoning focus.
CursorBench 4.0: 46.3% (vs Grok 4.6's 40.4% on the same benchmark version). DeepSWE v1.1 (high-effort variant): 71.0%. Vendor-reported (xAI launch page, x.ai/news/grok-4-7). No SWE-bench Verified/Pro score published; these are xAI's own comparison benchmarks, not yet formal anchors in the companion modelglass-coding registry.
Native Grok Bot integration and "improved work verification" per the launch page. Terminal-Bench 4.0: 38.0% (vs Grok 4.6's 20.3% on the same version) — a large jump on agentic terminal-task completion.
500K tokens reported — see architecture note above on verification confidence.
Standard variant. A "fast" variant exists at 2x price with 2x output speed (per xAI's own launch copy, same as grok-4-6) — not separately rated here, see the registry pricing entry's notes.
Same $2/$6 per-1M pricing as Grok 4.6 despite a larger base model and higher benchmark scores across the board — no price increase for the generational jump.
Operator guidance
xAI's own recommendation: "the most capable model we've built" for code, chat, and general purposes. On Artificial Analysis's GDPval-AA v2.1 knowledge-work leaderboard it scores 1,695 Elo (xhigh), above GPT-5.6 Sol (1,588, max) and Kimi K3 (1,524, max) on the same v2.1 scale (checked 2026-09-23). A reasonable default when routing for sustained, multi-hour agentic work where 4.6's lower price isn't the deciding factor.
Use cases
- Long-running agentic workflows needing careful self-verification over many hours
- General-purpose coding, terminal/agentic tasks, and knowledge work
- Document/presentation creation and professional knowledge-work simulation (GDPval, AA Briefcase task types)
Limitations
- Released one day before this entry — pricing/benchmark figures are from launch-day materials only, not independently re-verified over time yet.
- Capability ratings beyond the cited benchmark figures are qualitative, not single cited benchmark runs.