SpaceXAI’s Grok 4.7 ships at the same price as Grok 4.6
The coding and knowledge-work model costs $2 per million input tokens and $6 per million output, SpaceXAI says.
Share
1 source

SpaceXAI launched Grok 4.7, pitched as its strongest model for coding and knowledge work — longer on hard tasks, more self-checking, new safeguard stack — at the same price and speed as Grok 4.6: $2 per million input tokens and $6 per million output. Live today in Cursor, Grok Build, the Grok API, third-party coding harnesses, and model routers / cloud platforms. A fast variant doubles output speed at double price.
What changed vs 4.6: Larger base, longer RL on harder multi-hour tasks, better self-verification and long context, plus training to natively understand the Grok Bot harness for chat and general knowledge work. Company chart (attribute to xAI): CursorBench 4.0 46.3% (vs 4.6’s 40.4%); DeepSWE v1.1 71.0%* high effort; EEBench 64.0%; AA Briefcase v1.1 1,657; Terminal-Bench 4.0 38.0%; Harvey Legal Agent 19.6%; HealthBench Professional 56.7%. Docs/slides and multi-hour office work called out as improved vs 4.6 on GDPval / Briefcase.
Safety: New refusal / jailbreak stack. Dual-use cyber and bio: company says it leads utility on benign work and refusal on dangerous prompts; LatchBio biosafety 62.4%; HackerBench v0.3 allows 3.3% of risky dual-use prompts through. Select cyber partners get invite-only red-team access for defense research.
Price context: Same $2/$6 as 4.6. Table compares to GPT-5.6 Sol Max at $4/$20 and Fable 5.1 Max at $10/$50 — the “half the price of comparable models” line is that comparison, not a cut from prior Grok. “Twice as fast, at half the price of comparable models” is marketing framing vs named peers.
Who for: Cursor / agentic coding users and API knowledge-work pipelines who want Grok’s pricing with a stronger long-task / verification claim. Replaces “stay on 4.6” only if the bench deltas matter for their workloads.
Risks: Vendor benches rule the scoreboard until third parties reproduce. Safety numbers are self-reported. Fast tier cost doubles output — watch bill surprise.
Source: xAI / SpaceXAI news, Grok 4.7.
— Rowan Keel
Edition Sept. 21, 2026 · Archive shelf


