Model comparison
Grok 4.6 vs Grok 4.7
GrokCompare published specifications, vendor API list prices and measured Borg performance. Both models are included in Borg.
Grok 4.6
Grok 4.7
Included in Borg
At a glance
| Specification | Grok 4.6 | Grok 4.7 |
|---|---|---|
| Provider | xAI | xAI |
| Release | 2026-08-12 Vendor source | 2026-09-21 Vendor source |
| Context | 500,000 tokens Vendor source | 500,000 tokens Vendor source |
| Modalities | Input: text, image; output: text | Input: text, image; output: text |
| Reasoning / thinking modes | Yes / Not published | Yes / Not published |
| Open weights | No | No |
| Vendor API list price / 1M tokens | $2 input / $6 output Global rates for prompts below 200,000 input tokens. At 200,000 input tokens or more, all tokens use long-context rates: $4 input / $12 output / $1 cached input per 1M tokens. US regional inference costs 1.1× the global rate. | $2 input / $6 output Global rates for prompts below 200,000 input tokens. At 200,000 input tokens or more, all tokens use long-context rates: $4 input / $12 output / $1 cached input per 1M tokens. US regional inference costs 1.1× the global rate. |
| Borg | Included in Borg subscription | Included in Borg subscription |
xAIMeasured on Borg
Metrics are not directly comparable to vendor benchmark claims. Models served different requests; this is observed gateway traffic, not a controlled head-to-head test.
Time to first token (gateway proxy)
Output tokens per second (whole turn)
| Metric | Grok 4.6 | Grok 4.7 |
|---|---|---|
| Time to first token (gateway proxy) p50 | 3252 ms (3471 samples) | 1037 ms (1436 samples) |
| Tokens per second (end to end) p50 | 13.6 output tokens per second (32690 samples) | 33 output tokens per second (1467 samples) |
| Request success rate | Not published | Not published |
| Coding bench pass rate | Not published | Not published |
Measurement window, samples and methodology
The fleet reference is the median of published model medians, not a request-weighted result. Prompt lengths, reasoning settings and traffic differ between models.
Grok 4.6: Measured on Borg: SearchAtlas AI Gateway production request logs. 30-day window ending 2026-10-08; 34616 requests. Minimum 100 samples per published request metric.
Grok 4.7: Measured on Borg: SearchAtlas AI Gateway production request logs. 30-day window ending 2026-10-08; 1681 requests. Minimum 100 samples per published request metric.
Completion speed covers the whole turn including queueing and reasoning time; it is not a streaming tokens-per-second figure.
- Time to first token (gateway proxy): time until the first streamed response event reaches the gateway; our closest available measure of time to first token
- Tokens per second (end to end): tokens per second, end to end: provider-reported output tokens divided by the whole turn's elapsed time, including queueing and reasoning; not a streaming per-token rate
How to choose
Grok 4.6 is listed for multi-step codebase changes, whereas Grok 4.7 is listed for long-running agentic coding. Choosing between Grok 4.6 and Grok 4.7 requires checking published specifications against observed Borg traffic for your workflow.
Key differences
| Spec | Grok 4.6 | Grok 4.7 |
|---|---|---|
| Context window | 500,000 tokens Vendor source | 500,000 tokens Vendor source |
| Reasoning modes | Yes / Not published | Yes / Not published |
| Release date | 2026-08-12 Vendor source | 2026-09-21 Vendor source |
| API price input / 1M | 2 Global rates for prompts below 200,000 input tokens. At 200,000 input tokens or more, all tokens use long-context rates: $4 input / $12 output / $1 cached input per 1M tokens. US regional inference costs 1.1× the global rate. | 2 Global rates for prompts below 200,000 input tokens. At 200,000 input tokens or more, all tokens use long-context rates: $4 input / $12 output / $1 cached input per 1M tokens. US regional inference costs 1.1× the global rate. |
| API price output / 1M | 6 Global rates for prompts below 200,000 input tokens. At 200,000 input tokens or more, all tokens use long-context rates: $4 input / $12 output / $1 cached input per 1M tokens. US regional inference costs 1.1× the global rate. | 6 Global rates for prompts below 200,000 input tokens. At 200,000 input tokens or more, all tokens use long-context rates: $4 input / $12 output / $1 cached input per 1M tokens. US regional inference costs 1.1× the global rate. |
| Time to first token (gateway proxy) | 3252 ms (3471 samples) | 1037 ms (1436 samples) Lower median |
| Output tokens per second (whole turn) | 13.6 output tokens/s (whole turn) (32690 samples) | 33 output tokens/s (whole turn) (1467 samples) Higher measured rate |
| Request success | Not published | Not published |
| Coding bench | Not published | Not published |
FAQ
How do Grok 4.6 and Grok 4.7 compare for coding?
Grok 4.6 has a Borg coding benchmark of not published; Grok 4.7 has not published. The Grok 4.6–Grok 4.7 coding comparison cannot establish an overall winner from this small test alone.
How quickly do Grok 4.6 and Grok 4.7 start responding on Borg?
Borg median first response for Grok 4.6 is 3,252 ms; for Grok 4.7 it is 1,037 ms. The Grok 4.6–Grok 4.7 gateway medians are observations, not guaranteed latency for individual requests.
What are the API prices for Grok 4.6 and Grok 4.7?
The listed API rate for Grok 4.6 is $2 input / $6 output per 1M tokens; for Grok 4.7 it is $2 input / $6 output per 1M tokens. Neither Grok 4.6 nor Grok 4.7 has its Borg subscription charge represented by those provider API rates. Grok 4.6 vendor source · Grok 4.7 vendor source
Claude
Gemini