xAI Report:Grok 4.7 released, but performance still trails some competitors

Article

xAI released Grok 4.7: parameter scale increased to 2100000000000, while benchmarks still lag behind Claude

Dongcha Beating AI breaking news:xAI officially released Grok 4.7, with the company claiming “significant improvements” over Grok 4.6 at the same price and speed.Grok 4.7 previously underwent at least 5 delays and is now available on the Grok app, Cursor,Grok Build, and xAI API。

Grok 4.7 has reached a parameter scale of 2100000000000, an increase of Grok 4.6 over 1500000000000's 40%. Input and output pricing is Token 2 USD and 6 USD。xAI per million tokens, respectively. It also introduced supplemental SpaceX training data, including Starlink satellite telemetry data, manufacturing records, and engineering failure logs, aiming to improve the model's reasoning about hardware and physical systems.

In benchmark testing,Grok 4.7 ranked second in both GDPval and AA-Briefcase, with scores of 1695 and 1657, respectively, both behind Claude Fable 5.1's 1735 and 1678; it also ranked second in the electrical-engineering benchmark EEBench, behind only GPT-6 Astra。CursorBench 4.0 shows that Grok 4.7's task cost is higher than GPT-6 Astra and Claude Sonnet 5, while its overall performance still trails Fable 5.1.

Musk previously said that Grok 4.7 is roughly comparable to Claude Opus 5.0, but has not yet reached the level of the newer Opus 5.1, and disclosed that Grok 4.8、4.9 and Grok 5 are under development.

Original link https://m.theblockbeats.info/flash/368305

Key points

xAI Officially released on Monday Grok 4.7。
Grok 4.7 Previously underwent at least 5 delays and is now available on the Grok app, Cursor,Grok Build, and xAI API。
Grok 4.7 The parameter scale is 2100000000000, an increase of Grok 4.6 over 1500000000000's 40%。
Input and output pricing is Token 2 USD and 6 USD。
Grok 4.7 Ranked second in both GDPval and AA-Briefcase, with scores of 1695 and 1657。
Grok 4.7 Ranked second in EEBench, behind only GPT-6 Astra。
CursorBench 4.0 shows that Grok 4.7's task cost is higher than GPT-6 Astra and Claude Sonnet 5, while overall performance still trails Fable 5.1。
Musk once said that Grok 4.7 is roughly comparable to Claude Opus 5.0, but has not yet reached the level of Opus 5.1.
xAI Introduced supplemental SpaceX training data, including Starlink satellite telemetry data, manufacturing records, and engineering failure logs.

AI-assisted interpretation

The following is analysis, separate from reported facts. Verify important claims independently.

This is a new model release. Its parameter scale and training data have expanded, but multiple reported test results indicate that it still trails some Claude and GPT models.

Why it matters to readers

Readers can learn about Grok 4.7's launch status, pricing, scale, training-data sources, and benchmark performance.

Parameter scale, pricing, and benchmark scores can help newcomers make an initial comparison of different models' capabilities and usage costs.

Risks and unknowns

The data and rankings in the article are sourced from the report and have not been independently verified.
“Significant improvement” is the official wording; the text does not provide a specific magnitude of improvement.
The text does not specify the testing methods or sample scope for each benchmark.

xAI Report:Grok 4.7 released, but performance still trails some competitors

Article

AI-assisted interpretation

Why it matters to readers

Risks and unknowns

Related Developments

Related concepts

Keep following

Security Center