grok-4.6 is live at xAI’s own price, with 20% off on the GrokOfficial group
xAI (SpaceXAI)‘s new flagship, shipped 7 August. This generation did not change the base model — it reuses Grok 4.5’s 1.5T-parameter V9 foundation, with the entire gain coming from heavier supervised fine-tuning and reinforcement learning, so speed, token efficiency and list price all match Grok 4.5. The Artificial Analysis Intelligence Index (High tier) rises from 56 to 61, level with GPT-5.6 Sol Max, with top-tier results on GDPval-AA v2, AA-Briefcase and Harvey LAB. xAI reports markedly stronger self-testing and verification on long-running tasks.
Specs and integration: 500K context, text and image input, both /v1/chat/completions and /v1/responses, integrated exactly like the rest of the Grok 4.x line — just change the model name.
Billing runs in two context tiers, identical to xAI’s own:
- 0 – 200K tokens: $2.00 input, $6.00 output, $0.50 cached read (per 1M tokens)
- 200K – 512K tokens: $4.00 input, $12.00 output, $1.00 cached read (per 1M tokens)
GrokOfficial direct-relay group runs at a 0.8x multiplier (20% off the default group), with identical model behavior and call syntax — select it when creating a Token, no code changes. The discount stacks with top-up bonuses (10%–20%), bringing first-tier input as low as $1.33 and output to $4.00.
Existing grok-4.5 workloads get the same price and speed, so changing the model name is pure upside. Full benchmarks, capability matrix and code samples: Grok 4.6 launch notes.
← Back to Live Updates · 📚 Monthly archive