Breaking

SpaceXAI Launches Grok 4.6 for Long-Running Agents at $2/$6 per Million Tokens

August 12, 2026 at 12:32 EDT

Foundation Models
AI Agents
Software Dev & Coding

SpaceXAI released Grok 4.6 on August 12, positioning the frontier model for long-running agents, coding and complex knowledge work. It matches GPT-5.6 Sol on a broad intelligence index while starting at $2 per million input tokens and $6 per million output tokens.

Aug 12, 2026 · SpaceXAI

Grok 4.6 targets the long-running agent race

The new model reaches a 61 intelligence score while undercutting two frontier rivals on published API pricing.

AA Intelligence Index

500K

Context tokens

Input per 1M tokens

Output per 1M tokens

Published base output pricing

Each stacked block represents $2 per million output tokens.

Grok 4.6

$30

GPT-5.6 Sol

$50

Claude Fable 5

Pricing condition: Grok prompts above 200K tokens are billed at $4 input and $12 output per million tokens; its fast variant costs twice the base rate.

Training path

Longer supplemental run

→

Refreshed SFT trajectories

→

Problematic traces filtered

→

Agentic reinforcement learning

The target is sustained work across research, codebases, web development, computer-aided design and iterative product building.

Why developers care

Lower-cost frontier access

Strong professional-work scores and immediate Cursor integration make long codebase and prototype runs more affordable.

What remains uncertain

Benchmarks need field testing

Independent speed data is unavailable, token use is slightly above the peer median, and rivals still lead several coding tests.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.