Releases·Model Release
Grok 4.7 is now available on Amazon Bedrock
xAI's Grok 4.7 is now on Amazon Bedrock via cross-Region inference profiles, with a 500K token context window, four reasoning effort levels, and Responses, Chat Completions, and Converse API support. xAI describes it as its most capable coding and knowledge-work model, built on a new larger base model with a longer RL run weighted toward long-duration tasks and self-verification. Artificial Analysis's independent evaluation confirms gains over Grok 4.6 across its suite, but at roughly double the output tokens per task.
0 engagement2 sourcesMon, Sep 28, 2026, 12:00 AM
AI Overview
- Grok 4.7 on Bedrock offers a 500K token context window, four reasoning effort levels (default high), image input, and tool calling.
- xAI says Grok 4.7 is trained on a new, larger base model with a longer RL run weighted toward tasks taking many hours, improving self-verification.
- Artificial Analysis measures Grok 4.7 at Intelligence Index 46 vs Grok 4.6's 44, with roughly double the output tokens per task (~81k vs ~38k).
- The model is served via cross-Region inference profiles (us.xai.grok-4.7, global.xai.grok-4.7) through the Responses, Chat Completions, and Converse APIs.
- Bedrock deployment supports implicit prompt caching, Guardrails, structured outputs, invocation logging, and Standard, Priority, and Flex service tiers.
- xAI reports Grok 4.7 has a new safeguard stack, its best jailbreak resistance, and invite-only red-team access for selected cybersecurity partners.
- Bedrock access requires IAM permissions on inference profiles plus bedrock:CallWithBearerToken for bearer-token auth; production should use short-term IAM-minted tokens.
Related
Releases·Model Release
Build long-horizon agents with Grok 4.7 on Amazon Bedrock
An AWS post demonstrates building long-horizon tool-calling agents with SpaceXAI's Grok 4.7, newly available on Amazon Bedrock. Across 72 agent runs on three search-based tasks, the agent answered 67 correctly, with xhigh reasoning effort scoring 18 of 18 and low effort scoring 15 of 18 while using about an eighth of the output tokens. The post gives practitioners a tuning guide for the reasoning-effort trade-off and Bedrock integration specifics.
An AWS post demonstrates building long-horizon tool-calling agents with SpaceXAI's Grok 4.7, newly available on Amazon Bedrock. Across 72 agent runs on three search-based tasks, the agent answered 67 correctly, with xhigh reasoning effort scoring 18 of 18 and low effort scoring 15 of 18 while using about an eighth of the output tokens. The post gives practitioners a tuning guide for the reasoning-effort trade-off and Bedrock integration specifics.
Updated Tue, Oct 13, 2026, 03:52 PM — AWS published an evaluation showing reasoning-effort tuning: xhigh scored 18/18 on agent tasks, low scored 15/18 at ~1/8
0 engagement1 sourcerssTue, Oct 13, 2026, 03:52 PM
Social·Community
Actually positively surprised by Grok 4.7 EXTRA HIGH
A user who was unhappy with Grok 4.7 at release, having tested only medium and high reasoning effort, reports that EXTRA HIGH effort feels strictly better than Grok 4.6 high on their larger projects. They switched while trying to use up remaining monthly token allowance and note they can't tell whether xAI changed anything under the hood.
A user who was unhappy with Grok 4.7 at release, having tested only medium and high reasoning effort, reports that EXTRA HIGH effort feels strictly better than Grok 4.6 high on their larger projects. They switched while trying to use up remaining monthly token allowance and note they can't tell whether xAI changed anything under the hood.
35 engagement1 sourceredditSun, Oct 11, 2026, 10:53 AM
Releases·Model Release
xAI announces Grok 4.6 and Grok 4.7 release schedule
xAI has announced that Grok 4.6 will be released in two weeks and Grok 4.7 in four weeks, as of July 24, 2026. The rapid iteration suggests significant improvements in reasoning and capabilities, likely impacting developers and researchers using the Grok platform.
xAI has announced that Grok 4.6 will be released in two weeks and Grok 4.7 in four weeks, as of July 24, 2026. The rapid iteration suggests significant improvements in reasoning and capabilities, likely impacting developers and researchers using the Grok platform.
334 engagement1 sourceredditFri, Jul 24, 2026, 07:43 PM
Social·Community
Cursor users weigh leaving as Grok 4.6/4.7 lag Opus 5.5 and GPT 6.1 on price and token efficiency
A Reddit thread argues xAI's Grok 4.6 and 4.7, as offered through Cursor, now look overpriced and slow after Anthropic's Opus 5.5 and OpenAI's 6.1 cut prices and improved token efficiency. The poster questions renewing Cursor subscriptions given stricter API-priced limits than OpenAI/Anthropic subscription plans. The thread is user sentiment, not an announcement.
A Reddit thread argues xAI's Grok 4.6 and 4.7, as offered through Cursor, now look overpriced and slow after Anthropic's Opus 5.5 and OpenAI's 6.1 cut prices and improved token efficiency. The poster questions renewing Cursor subscriptions given stricter API-priced limits than OpenAI/Anthropic subscription plans. The thread is user sentiment, not an announcement.
175 engagement2 sourcesredditTue, Sep 29, 2026, 09:01 PM
Social·Community
Grok 4.7 is about 2.5 times as expensive as 4.6
A Cursor Ultra user compared Grok 4.7 and 4.6 on the same coding task and found 4.7 costs roughly 2.5 times as much, matching Artificialanalysis.ai measurements ($8.82 vs $3.53 per task). Given only ~10% benchmark gains, they plan to keep using 4.6 for most tasks.
A Cursor Ultra user compared Grok 4.7 and 4.6 on the same coding task and found 4.7 costs roughly 2.5 times as much, matching Artificialanalysis.ai measurements ($8.82 vs $3.53 per task). Given only ~10% benchmark gains, they plan to keep using 4.6 for most tasks.
137 engagement2 sourcesredditMon, Sep 21, 2026, 11:23 PM