Grok 4.6 matches GPT-5.6 and Opus 5 on many benchmarks, per @kimmonismus, while xAI readies a 2.1T-parameter Grok 4.7 for release in weeks. The 1.5T model ships with Cursor integration, a fast turnaround for a frontier lab.
Key facts
- Grok 4.6 is a 1.5T-parameter model
- Grok 4.7 will be 2.1T parameters, releasing in weeks
- 4.6 matches GPT-5.6, Opus 5 on many benchmarks
- 4.7 better than 4.6 except slower serving
- Kimi k3.1 and GLM-5.3 (Flash) add competitive pressure
xAI's Grok 4.6, a 1.5T-parameter model, is delivering benchmark parity with the top frontier models from OpenAI, Anthropic and Mistral. According to @kimmonismus, the model "performs on par with GPT-5.6, Opus 5, and even, in certain benchmarks, Fable 5" — a notable result for a model that is not xAI's largest. The same source calls it "excellent value for money," a claim the company has not yet backed with pricing disclosures.
The 2.1T follow-up
The key context: Grok 4.6 is the smaller sibling. "Grok 4.7 will be the 2.1T model released a few weeks later," @kimmonismus writes, describing it as "better than 4.6 in every way, except slightly slower to serve, albeit with even better token efficiency." That trade-off — quality per token vs. latency — is the classic frontier-model calculus, and it suggests xAI is optimizing 4.7 for compute efficiency rather than raw speed.
The 4.6 release also bundles Cursor integration, a move that puts xAI's models directly into the developer workflow where OpenAI's Codex and Anthropic's Claude Code currently compete. Shipping a frontier-adjacent model with an IDE integration in short order is a distribution play as much as a model play.
Pressure on the incumbents
The competitive window is tightening. "Kimi k3.1 is about to be released, GLM-5.3 (Flash) is currently demonstrating how good smaller models can be," @kimmonismus notes, "and thus the pressure on OpenAI and Anthropic is increasing." The pattern across the last quarter: smaller, cheaper models from Chinese labs and xAI are compressing the price-performance gap faster than the frontier leaders are extending it.
None of this is a formal benchmark disclosure. xAI has not published third-party eval results for 4.6, and the source's claims about Fable 5 parity in "certain benchmarks" are unspecified. The company did not disclose parameter counts for 4.6 beyond the 1.5T figure referenced in the tweet, nor did it confirm the 2.1T size for 4.7.
What this means for buyers
For teams choosing a model today, the calculus is shifting: a 1.5T model that matches the big three on many tasks, at presumably lower cost, changes the default. The real question is whether Grok 4.7's token efficiency — better than 4.6's, per the source — translates to a price advantage that forces OpenAI and Anthropic to respond on pricing rather than just capability.
Watch for xAI's pricing page update when 4.7 drops, and whether OpenAI or Anthropic announce price cuts within the following two weeks. The window for the incumbents to defend their per-token margins is closing.
What to watch
Watch for the Grok 4.7 release window — the source says weeks, not months. When it ships, check xAI's pricing per million tokens against OpenAI's GPT-5.6 and Anthropic's Opus 5. A price cut from either incumbent within two weeks of the 4.7 launch would confirm margin pressure.
