Grok 4.6 is now available in ThinkReview
Grok 4.6 is now available in ThinkReview
We're excited to announce that Grok 4.6 is now available in ThinkReview. SpaceXAI's Grok 4.6 builds on Grok 4.5 with a sharper focus on long-running agents, sustained engineering work, and more ambitious interactive and visual projects. The result, in our reviews, is impressive finding quality — the kind of careful, mechanism-level feedback that holds up on real pull requests. You can use it today on GitHub, GitLab, Azure DevOps, and Bitbucket.
What is Grok 4.6?
Grok 4.6 is SpaceXAI's latest frontier model. The launch positions it less as a chat upgrade and more as a model that stays with hard, multi-step work — researching a topic, analyzing a codebase, or turning an idea into a working artifact without losing the thread.
Highlights from SpaceXAI:
- Built for long-running agents — Trained to stay on complex tasks across many steps, including software engineering and knowledge work.
- Frontier intelligence — Matches GPT-5.6 Sol on the Artificial Analysis Intelligence Index (composite of nine benchmarks).
- Stronger coding evals than Grok 4.5 — Clear lifts on CursorBench 3.2, DeepSWE 1.1, FrontierCode 1.1, and APEX-SWE.
- Self-testing on longer trajectories — SpaceXAI reports more verification before the model moves on — useful when a review should check its own claims.
- Same aggressive pricing — $2 / $6 per million input/output tokens, with a faster variant at twice that price.
SpaceXAI also describes a longer supplemental training run than 4.5, regenerated SFT trajectories, and agentic RL across general coding plus domain environments such as kernel optimization and web development. Grok 4.6 is available in Cursor, Grok Build, the SpaceXAI API, and partners including OpenRouter — and now inside ThinkReview.
Why Grok 4.6 is impressive for code reviews
Pull request review is a long-horizon job. The model has to read a diff, pull related files, keep the surrounding architecture in mind, and write findings a human reviewer would actually act on. That is the workload Grok 4.6 is tuned for.
- Higher-signal first passes — Stronger coding and agent benchmarks than Grok 4.5 show up as sharper comments on bugs, regressions, and API contract breaks — not more noise.
- Stays with messy, multi-file PRs — Long-running agent training helps when a change spans services, configs, and tests instead of a single hunk.
- Checks itself before commenting — More self-testing and verification means fewer confident-but-wrong notes on race conditions, auth gaps, and edge-case failures.
- Frontier quality at Lite economics — Sol-class intelligence scores at $2 / $6 pricing make Grok 4.6 a practical default when you want impressive review quality without flagship spend on every MR.
- Comfortable in code-centric loops — Trained alongside Cursor / Grok Build workflows, so it fits ThinkReview's fetch-context, reason, comment cycle.
Selected Grok 4.6 High vs Grok 4.5 High results from SpaceXAI's announcement:
- AA Intelligence Index — 61 vs 56 (tied with GPT-5.6 Sol Max)
- CursorBench v3.2 — 69.9% vs 66.7%
- DeepSWE v1.1 — 65.9% vs 54%
- FrontierCode v1.1 (Extended) — 61.3% vs 56.6%
- APEX-Agents — 57.5% vs 47.1%
- APEX-SWE — 56.4% vs 53.6%
Benchmarks are not a substitute for your team's review bar. Treat severity labels as input, and pair Grok 4.6 with repository-level context when the real bug lives outside the hunk.
How to use Grok 4.6 in ThinkReview
- Open ThinkReview settings — Click the extension icon and go to Settings or Model selection.
- Choose Grok 4.6 — Select it from the model dropdown for your reviews.
- Run a review — Open any pull request or merge request and start ThinkReview with your chosen model.
Grok 4.6 is available on Lite (and plans that include Lite models). Manage your catalog in Model Selection on the ThinkReview portal.
Where it works
Same as always: GitHub, GitLab, Azure DevOps, and Bitbucket Cloud. One extension, one workflow — now with SpaceXAI's Grok 4.6 for impressive-quality AI reviews.
If you try Grok 4.6 on real PRs and have feedback, we'd love to hear it — open an issue on GitHub or reach out via thinkreview.dev.
Ready to try Grok 4.6 on your next PR? Install ThinkReview or manage your models in the portal.
Model details reference SpaceXAI's Grok 4.6 announcement.