{
  "@context": "https://schema.org",
  "@id": "https://aggr.aymericbeaumet.com/items/x-ai/2025-09-19-grok-4-fast/#webpage",
  "@type": [
    "WebPage",
    "ArchiveComponent"
  ],
  "articleBody": "![Abstract digital hummingbird](https://media.x.ai/cdn-cgi/image/fit=scale-down,onerror=redirect,width=6016,f=auto/v1/website/grok-4-fast-8c8d3e84.webp)\n\nWe're thrilled to present Grok 4 Fast, our latest advancement in cost-efficient reasoning models. Built on xAI’s learnings from Grok 4, Grok 4 Fast delivers frontier-level performance across Enterprise and Consumer domains—with exceptional token efficiency. This model pushes the boundaries for smaller and faster AI, making high-quality reasoning accessible to more users and developers. Grok 4 Fast features state-of-the-art (SOTA) cost-efficiency, cutting-edge web and X search capabilities, a 2M token context window, and a unified architecture that blends `reasoning` and `non-reasoning` modes in one model.\n\n## [Advancing Cost-Efficient Intelligence](https://x.ai/news/grok-4-fast#advancing-cost-efficient-intelligence)\n\nGrok 4 Fast sets a new frontier in cost-efficient intelligence, outperforming Grok 3 Mini across reasoning benchmarks while slashing token costs.\n\n| Benchmark pass@1 | Grok 4 Fast | Grok 4 | Grok 3 Mini (High) | GPT-5 (High) | GPT-5 Mini (High) |\n| ---------------- | ----------- | ------ | ------------------ | ------------ | ----------------- |\n| 85.7%            | 87.5%       | 79.0%  | 85.7%              | 82.3%        |                   |\n| 92.0%            | 91.7%       | 83.0%  | 94.6%              | 91.1%        |                   |\n| 93.3%            | 90.0%       | 74.0%  | 93.3%              | 87.8%        |                   |\n| 20.0%            | 25.4%       | 11.0%  | 24.8%              | 16.7%        |                   |\n| 80.0%            | 79.0%       | 70.0%  | 86.8%              | 77.4%        |                   |\n\nWe used large-scale reinforcement learning to maximize the intelligence density of Grok 4 Fast. In our evaluations, Grok 4 Fast achieves comparable performance to Grok 4 on benchmarks while using 40% fewer thinking tokens on average.\n\nThis 40% increase in Grok 4 Fast's token efficiency, combined with a significantly lower price per token, results in a 98% reduction in price to achieve the same performance on frontier benchmarks as Grok 4. As verified by an independent review from Artificial Analysis, Grok 4 Fast exhibits a state-of-the-art (SOTA) price-to-intelligence ratio compared to other publicly available models on the Artificial Analysis Intelligence Index.\n\n\\*All Claude models were benchmarked with Extended Thinking.\n\n## [Native Tool Use with SOTA Search](https://x.ai/news/grok-4-fast#native-tool-use-with-sota-search)\n\nGrok 4 Fast was trained end-to-end with tool-use reinforcement learning (RL). It excels at deciding when to invoke tools like code execution or web browsing.\n\nFor instance, Grok 4 Fast exhibits frontier agentic search capabilities, seamlessly browsing the web and X to augment queries with real-time data. It hops through links, ingests media (including images and videos on X), and synthesizes findings at light speed.\n\n| Benchmark pass@1 | Grok 4 Fast | Grok 4 | Grok 3 (No Reasoning) |\n| ---------------- | ----------- | ------ | --------------------- |\n| 44.9%            | 43.0%       | —      |                       |\n| 95.0%            | 94.0%       | 82.0%  |                       |\n| 66.0%            | 58.0%       | 37.0%  |                       |\n| 51.2%            | 45.0%       | 10.8%  |                       |\n| 74.0%            | 66.0%       | 27.0%  |                       |\n| 58.0%            | 53.2%       | 20.8%  |                       |\n\n\\*X Browse is an internal benchmark evaluating agent's multihop search and browsing capabilities on X.\n\n## [Frontier of General Post-training](https://x.ai/news/grok-4-fast#frontier-of-general-post-training)\n\nGrok 4 Fast also establishes a new cost-effective frontier on general domain. We are excited to share Grok 4 Fast’s result on LMArena, where it has been privately battle-testing on the Search and Text Arenas.\n\nIn LMArena's [Search Arena](https://lmarena.ai/leaderboard/search), `grok-4-fast-search` (code name: `menlo`) claims ***#1*** with ***1163*** Elo — a commanding margin of ***17*** over `o3-search`. Its superior reasoning efficiency and intelligence density enable it to surpass much larger models on real-world, search-related tasks.\n\n![Artificial Intelligence Search Arena](https://media.x.ai/cdn-cgi/image/fit=scale-down,onerror=redirect,width=6016,f=auto/v1/website/search_arena-c5105183.webp)\n\nIn LMArena's [Text Arena](https://lmarena.ai/leaderboard/text), `grok-4-fast` (code name: `tahoe`) ranks ***#8***, performing on par with `grok-4-0709` and highlighting its remarkable intelligence density. Notably, it significantly outperforms peers in its weight class, where all comparable size models rank ***18th*** or below.\n\n![Artificial Intelligence Text Arena](https://media.x.ai/cdn-cgi/image/fit=scale-down,onerror=redirect,width=6016,f=auto/v1/website/text_arena-fcc81749.webp)\n\nHere are some examples of Grok 4 Fast in action:\n\nWhat is the maximum number of experience points possible in Path of Exile 2?\n\n## [Unified Model: Reasoning and Non-Reasoning](https://x.ai/news/grok-4-fast#unified-model-reasoning-and-non-reasoning)\n\nPreviously, separate reasoning modes required distinct models. Grok 4 Fast introduces a unified architecture where `reasoning` (long chain-of-thought) and `non-reasoning` (quick responses) are handled by the same model weights, steered via system prompts. This unification reduces end-to-end latency as well as token costs, making Grok 4 Fast ideal for real-time applications.\n\nIn [grok.com](http://grok.com/?q=What%20is%20the%20answer%20to%20life,%20the%20universe,%20and%20everything?&m=7), this results in smooth transitions: responding instantly for simple queries or engaging in extended reasoning for complex ones. In the xAI API, developers can fine-tune this behavior, optimizing for speed or depth.\n\nGrok 4 Fast is available now for all users. In `Fast` and `Auto` modes, you will see a significant improvement in search and information seeking queries. Additionally, difficult queries in `Auto` mode will use Grok 4 Fast, which will provide a much faster experience without loss of quality. For the first time, all users, including free users, will have access to our latest model without restrictions, marking a step toward democratizing advanced AI.\n\n## [Grok 4 Fast on OpenRouter, Vercel AI Gateway, and the xAI API](https://x.ai/news/grok-4-fast#grok-4-fast-on-openrouter-vercel-ai-gateway-and-the-xai-api)\n\nWe're rolling out Grok 4 Fast as two models: `grok-4-fast-reasoning` and `grok-4-fast-non-reasoning`, each with a 2M token context window. This allows developers to tune the amount of test-time compute applied to their use cases.\n\n`grok-4-fast-reasoning` and `grok-4-fast-non-reasoning` are generally available via the [xAI API](https://x.ai/api) according to the following pricing:\n\n| Token Type          | <128k tokens | ≥128k tokens |\n| ------------------- | ------------ | ------------ |\n| Input tokens        | $0.20 / 1M   | $0.40 / 1M   |\n| Output tokens       | $0.50 / 1M   | $1.00 / 1M   |\n| Cached input tokens | $0.05 / 1M   |              |\n\n## [What's Next](https://x.ai/news/grok-4-fast#whats-next)\n\nWe will continuously ship model improvements to Grok 4 Fast based on your feedback on [x.com](http://x.com). Stay tuned for further integrations, including enhanced multimodal capabilities and agentic features.\n\nRead the Grok 4 Fast model card [here](https://data.x.ai/2025-09-19-grok-4-fast-model-card.pdf).\n\nThat's all for now - so long, and thanks for all the fish!\n",
  "authors": [],
  "captured_at": "2026-10-19T01:26:53.998049939+00:00",
  "category": "ai",
  "content_html": "<p><picture class=\"article-picture article-picture-fallback\" style=\"--image-ratio:16 / 9\"><img src=\"https://media.x.ai/cdn-cgi/image/fit=scale-down,onerror=redirect,width=6016,f=auto/v1/website/grok-4-fast-8c8d3e84.webp\" alt=\"Abstract digital hummingbird\" loading=\"eager\" decoding=\"async\" fetchpriority=\"high\" referrerpolicy=\"no-referrer\" class=\"progressive-image\"></picture></p>\n<p>We're thrilled to present Grok 4 Fast, our latest advancement in cost-efficient reasoning models. Built on xAI’s learnings from Grok 4, Grok 4 Fast delivers frontier-level performance across Enterprise and Consumer domains—with exceptional token efficiency. This model pushes the boundaries for smaller and faster AI, making high-quality reasoning accessible to more users and developers. Grok 4 Fast features state-of-the-art (SOTA) cost-efficiency, cutting-edge web and X search capabilities, a 2M token context window, and a unified architecture that blends <code>reasoning</code> and <code>non-reasoning</code> modes in one model.</p>\n<h2><a target=\"_blank\" rel=\"noopener noreferrer\" href=\"https://x.ai/news/grok-4-fast#advancing-cost-efficient-intelligence\">Advancing Cost-Efficient Intelligence</a></h2>\n<p>Grok 4 Fast sets a new frontier in cost-efficient intelligence, outperforming Grok 3 Mini across reasoning benchmarks while slashing token costs.</p>\n<table>\n<thead>\n<tr>\n<th>Benchmark pass@1</th>\n<th>Grok 4 Fast</th>\n<th>Grok 4</th>\n<th>Grok 3 Mini (High)</th>\n<th>GPT-5 (High)</th>\n<th>GPT-5 Mini (High)</th>\n</tr>\n</thead>\n<tbody>\n<tr>\n<td>85.7%</td>\n<td>87.5%</td>\n<td>79.0%</td>\n<td>85.7%</td>\n<td>82.3%</td>\n<td></td>\n</tr>\n<tr>\n<td>92.0%</td>\n<td>91.7%</td>\n<td>83.0%</td>\n<td>94.6%</td>\n<td>91.1%</td>\n<td></td>\n</tr>\n<tr>\n<td>93.3%</td>\n<td>90.0%</td>\n<td>74.0%</td>\n<td>93.3%</td>\n<td>87.8%</td>\n<td></td>\n</tr>\n<tr>\n<td>20.0%</td>\n<td>25.4%</td>\n<td>11.0%</td>\n<td>24.8%</td>\n<td>16.7%</td>\n<td></td>\n</tr>\n<tr>\n<td>80.0%</td>\n<td>79.0%</td>\n<td>70.0%</td>\n<td>86.8%</td>\n<td>77.4%</td>\n<td></td>\n</tr>\n</tbody>\n</table>\n<p>We used large-scale reinforcement learning to maximize the intelligence density of Grok 4 Fast. In our evaluations, Grok 4 Fast achieves comparable performance to Grok 4 on benchmarks while using 40% fewer thinking tokens on average.</p>\n<p>This 40% increase in Grok 4 Fast's token efficiency, combined with a significantly lower price per token, results in a 98% reduction in price to achieve the same performance on frontier benchmarks as Grok 4. As verified by an independent review from Artificial Analysis, Grok 4 Fast exhibits a state-of-the-art (SOTA) price-to-intelligence ratio compared to other publicly available models on the Artificial Analysis Intelligence Index.</p>\n<p>*All Claude models were benchmarked with Extended Thinking.</p>\n<h2><a target=\"_blank\" rel=\"noopener noreferrer\" href=\"https://x.ai/news/grok-4-fast#native-tool-use-with-sota-search\">Native Tool Use with SOTA Search</a></h2>\n<p>Grok 4 Fast was trained end-to-end with tool-use reinforcement learning (RL). It excels at deciding when to invoke tools like code execution or web browsing.</p>\n<p>For instance, Grok 4 Fast exhibits frontier agentic search capabilities, seamlessly browsing the web and X to augment queries with real-time data. It hops through links, ingests media (including images and videos on X), and synthesizes findings at light speed.</p>\n<table>\n<thead>\n<tr>\n<th>Benchmark pass@1</th>\n<th>Grok 4 Fast</th>\n<th>Grok 4</th>\n<th>Grok 3 (No Reasoning)</th>\n</tr>\n</thead>\n<tbody>\n<tr>\n<td>44.9%</td>\n<td>43.0%</td>\n<td>—</td>\n<td></td>\n</tr>\n<tr>\n<td>95.0%</td>\n<td>94.0%</td>\n<td>82.0%</td>\n<td></td>\n</tr>\n<tr>\n<td>66.0%</td>\n<td>58.0%</td>\n<td>37.0%</td>\n<td></td>\n</tr>\n<tr>\n<td>51.2%</td>\n<td>45.0%</td>\n<td>10.8%</td>\n<td></td>\n</tr>\n<tr>\n<td>74.0%</td>\n<td>66.0%</td>\n<td>27.0%</td>\n<td></td>\n</tr>\n<tr>\n<td>58.0%</td>\n<td>53.2%</td>\n<td>20.8%</td>\n<td></td>\n</tr>\n</tbody>\n</table>\n<p>*X Browse is an internal benchmark evaluating agent's multihop search and browsing capabilities on X.</p>\n<h2><a target=\"_blank\" rel=\"noopener noreferrer\" href=\"https://x.ai/news/grok-4-fast#frontier-of-general-post-training\">Frontier of General Post-training</a></h2>\n<p>Grok 4 Fast also establishes a new cost-effective frontier on general domain. We are excited to share Grok 4 Fast’s result on LMArena, where it has been privately battle-testing on the Search and Text Arenas.</p>\n<p>In LMArena's <a target=\"_blank\" rel=\"noopener noreferrer\" href=\"https://lmarena.ai/leaderboard/search\">Search Arena</a>, <code>grok-4-fast-search</code> (code name: <code>menlo</code>) claims <em><strong>#1</strong></em> with <em><strong>1163</strong></em> Elo — a commanding margin of <em><strong>17</strong></em> over <code>o3-search</code>. Its superior reasoning efficiency and intelligence density enable it to surpass much larger models on real-world, search-related tasks.</p>\n<p><picture class=\"article-picture article-picture-fallback\" style=\"--image-ratio:16 / 9\"><img src=\"https://media.x.ai/cdn-cgi/image/fit=scale-down,onerror=redirect,width=6016,f=auto/v1/website/search_arena-c5105183.webp\" alt=\"Artificial Intelligence Search Arena\" loading=\"lazy\" decoding=\"async\" fetchpriority=\"low\" referrerpolicy=\"no-referrer\" class=\"progressive-image\"></picture></p>\n<p>In LMArena's <a target=\"_blank\" rel=\"noopener noreferrer\" href=\"https://lmarena.ai/leaderboard/text\">Text Arena</a>, <code>grok-4-fast</code> (code name: <code>tahoe</code>) ranks <em><strong>#8</strong></em>, performing on par with <code>grok-4-0709</code> and highlighting its remarkable intelligence density. Notably, it significantly outperforms peers in its weight class, where all comparable size models rank <em><strong>18th</strong></em> or below.</p>\n<p><picture class=\"article-picture article-picture-fallback\" style=\"--image-ratio:16 / 9\"><img src=\"https://media.x.ai/cdn-cgi/image/fit=scale-down,onerror=redirect,width=6016,f=auto/v1/website/text_arena-fcc81749.webp\" alt=\"Artificial Intelligence Text Arena\" loading=\"lazy\" decoding=\"async\" fetchpriority=\"low\" referrerpolicy=\"no-referrer\" class=\"progressive-image\"></picture></p>\n<p>Here are some examples of Grok 4 Fast in action:</p>\n<p>What is the maximum number of experience points possible in Path of Exile 2?</p>\n<h2><a target=\"_blank\" rel=\"noopener noreferrer\" href=\"https://x.ai/news/grok-4-fast#unified-model-reasoning-and-non-reasoning\">Unified Model: Reasoning and Non-Reasoning</a></h2>\n<p>Previously, separate reasoning modes required distinct models. Grok 4 Fast introduces a unified architecture where <code>reasoning</code> (long chain-of-thought) and <code>non-reasoning</code> (quick responses) are handled by the same model weights, steered via system prompts. This unification reduces end-to-end latency as well as token costs, making Grok 4 Fast ideal for real-time applications.</p>\n<p>In <a target=\"_blank\" rel=\"noopener noreferrer\" href=\"http://grok.com/?q=What%20is%20the%20answer%20to%20life,%20the%20universe,%20and%20everything?&amp;m=7\">grok.com</a>, this results in smooth transitions: responding instantly for simple queries or engaging in extended reasoning for complex ones. In the xAI API, developers can fine-tune this behavior, optimizing for speed or depth.</p>\n<p>Grok 4 Fast is available now for all users. In <code>Fast</code> and <code>Auto</code> modes, you will see a significant improvement in search and information seeking queries. Additionally, difficult queries in <code>Auto</code> mode will use Grok 4 Fast, which will provide a much faster experience without loss of quality. For the first time, all users, including free users, will have access to our latest model without restrictions, marking a step toward democratizing advanced AI.</p>\n<h2><a target=\"_blank\" rel=\"noopener noreferrer\" href=\"https://x.ai/news/grok-4-fast#grok-4-fast-on-openrouter-vercel-ai-gateway-and-the-xai-api\">Grok 4 Fast on OpenRouter, Vercel AI Gateway, and the xAI API</a></h2>\n<p>We're rolling out Grok 4 Fast as two models: <code>grok-4-fast-reasoning</code> and <code>grok-4-fast-non-reasoning</code>, each with a 2M token context window. This allows developers to tune the amount of test-time compute applied to their use cases.</p>\n<p><code>grok-4-fast-reasoning</code> and <code>grok-4-fast-non-reasoning</code> are generally available via the <a target=\"_blank\" rel=\"noopener noreferrer\" href=\"https://x.ai/api\">xAI API</a> according to the following pricing:</p>\n<table>\n<thead>\n<tr>\n<th>Token Type</th>\n<th>&lt;128k tokens</th>\n<th>≥128k tokens</th>\n</tr>\n</thead>\n<tbody>\n<tr>\n<td>Input tokens</td>\n<td>$0.20 / 1M</td>\n<td>$0.40 / 1M</td>\n</tr>\n<tr>\n<td>Output tokens</td>\n<td>$0.50 / 1M</td>\n<td>$1.00 / 1M</td>\n</tr>\n<tr>\n<td>Cached input tokens</td>\n<td>$0.05 / 1M</td>\n<td></td>\n</tr>\n</tbody>\n</table>\n<h2><a target=\"_blank\" rel=\"noopener noreferrer\" href=\"https://x.ai/news/grok-4-fast#whats-next\">What's Next</a></h2>\n<p>We will continuously ship model improvements to Grok 4 Fast based on your feedback on <a target=\"_blank\" rel=\"noopener noreferrer\" href=\"http://x.com\">x.com</a>. Stay tuned for further integrations, including enhanced multimodal capabilities and agentic features.</p>\n<p>Read the Grok 4 Fast model card <a target=\"_blank\" rel=\"noopener noreferrer\" href=\"https://data.x.ai/2025-09-19-grok-4-fast-model-card.pdf\">here</a>.</p>\n<p>That's all for now - so long, and thanks for all the fish!</p>\n",
  "content_markdown": "![Abstract digital hummingbird](https://media.x.ai/cdn-cgi/image/fit=scale-down,onerror=redirect,width=6016,f=auto/v1/website/grok-4-fast-8c8d3e84.webp)\n\nWe're thrilled to present Grok 4 Fast, our latest advancement in cost-efficient reasoning models. Built on xAI’s learnings from Grok 4, Grok 4 Fast delivers frontier-level performance across Enterprise and Consumer domains—with exceptional token efficiency. This model pushes the boundaries for smaller and faster AI, making high-quality reasoning accessible to more users and developers. Grok 4 Fast features state-of-the-art (SOTA) cost-efficiency, cutting-edge web and X search capabilities, a 2M token context window, and a unified architecture that blends `reasoning` and `non-reasoning` modes in one model.\n\n## [Advancing Cost-Efficient Intelligence](https://x.ai/news/grok-4-fast#advancing-cost-efficient-intelligence)\n\nGrok 4 Fast sets a new frontier in cost-efficient intelligence, outperforming Grok 3 Mini across reasoning benchmarks while slashing token costs.\n\n| Benchmark pass@1 | Grok 4 Fast | Grok 4 | Grok 3 Mini (High) | GPT-5 (High) | GPT-5 Mini (High) |\n| ---------------- | ----------- | ------ | ------------------ | ------------ | ----------------- |\n| 85.7%            | 87.5%       | 79.0%  | 85.7%              | 82.3%        |                   |\n| 92.0%            | 91.7%       | 83.0%  | 94.6%              | 91.1%        |                   |\n| 93.3%            | 90.0%       | 74.0%  | 93.3%              | 87.8%        |                   |\n| 20.0%            | 25.4%       | 11.0%  | 24.8%              | 16.7%        |                   |\n| 80.0%            | 79.0%       | 70.0%  | 86.8%              | 77.4%        |                   |\n\nWe used large-scale reinforcement learning to maximize the intelligence density of Grok 4 Fast. In our evaluations, Grok 4 Fast achieves comparable performance to Grok 4 on benchmarks while using 40% fewer thinking tokens on average.\n\nThis 40% increase in Grok 4 Fast's token efficiency, combined with a significantly lower price per token, results in a 98% reduction in price to achieve the same performance on frontier benchmarks as Grok 4. As verified by an independent review from Artificial Analysis, Grok 4 Fast exhibits a state-of-the-art (SOTA) price-to-intelligence ratio compared to other publicly available models on the Artificial Analysis Intelligence Index.\n\n\\*All Claude models were benchmarked with Extended Thinking.\n\n## [Native Tool Use with SOTA Search](https://x.ai/news/grok-4-fast#native-tool-use-with-sota-search)\n\nGrok 4 Fast was trained end-to-end with tool-use reinforcement learning (RL). It excels at deciding when to invoke tools like code execution or web browsing.\n\nFor instance, Grok 4 Fast exhibits frontier agentic search capabilities, seamlessly browsing the web and X to augment queries with real-time data. It hops through links, ingests media (including images and videos on X), and synthesizes findings at light speed.\n\n| Benchmark pass@1 | Grok 4 Fast | Grok 4 | Grok 3 (No Reasoning) |\n| ---------------- | ----------- | ------ | --------------------- |\n| 44.9%            | 43.0%       | —      |                       |\n| 95.0%            | 94.0%       | 82.0%  |                       |\n| 66.0%            | 58.0%       | 37.0%  |                       |\n| 51.2%            | 45.0%       | 10.8%  |                       |\n| 74.0%            | 66.0%       | 27.0%  |                       |\n| 58.0%            | 53.2%       | 20.8%  |                       |\n\n\\*X Browse is an internal benchmark evaluating agent's multihop search and browsing capabilities on X.\n\n## [Frontier of General Post-training](https://x.ai/news/grok-4-fast#frontier-of-general-post-training)\n\nGrok 4 Fast also establishes a new cost-effective frontier on general domain. We are excited to share Grok 4 Fast’s result on LMArena, where it has been privately battle-testing on the Search and Text Arenas.\n\nIn LMArena's [Search Arena](https://lmarena.ai/leaderboard/search), `grok-4-fast-search` (code name: `menlo`) claims ***#1*** with ***1163*** Elo — a commanding margin of ***17*** over `o3-search`. Its superior reasoning efficiency and intelligence density enable it to surpass much larger models on real-world, search-related tasks.\n\n![Artificial Intelligence Search Arena](https://media.x.ai/cdn-cgi/image/fit=scale-down,onerror=redirect,width=6016,f=auto/v1/website/search_arena-c5105183.webp)\n\nIn LMArena's [Text Arena](https://lmarena.ai/leaderboard/text), `grok-4-fast` (code name: `tahoe`) ranks ***#8***, performing on par with `grok-4-0709` and highlighting its remarkable intelligence density. Notably, it significantly outperforms peers in its weight class, where all comparable size models rank ***18th*** or below.\n\n![Artificial Intelligence Text Arena](https://media.x.ai/cdn-cgi/image/fit=scale-down,onerror=redirect,width=6016,f=auto/v1/website/text_arena-fcc81749.webp)\n\nHere are some examples of Grok 4 Fast in action:\n\nWhat is the maximum number of experience points possible in Path of Exile 2?\n\n## [Unified Model: Reasoning and Non-Reasoning](https://x.ai/news/grok-4-fast#unified-model-reasoning-and-non-reasoning)\n\nPreviously, separate reasoning modes required distinct models. Grok 4 Fast introduces a unified architecture where `reasoning` (long chain-of-thought) and `non-reasoning` (quick responses) are handled by the same model weights, steered via system prompts. This unification reduces end-to-end latency as well as token costs, making Grok 4 Fast ideal for real-time applications.\n\nIn [grok.com](http://grok.com/?q=What%20is%20the%20answer%20to%20life,%20the%20universe,%20and%20everything?&m=7), this results in smooth transitions: responding instantly for simple queries or engaging in extended reasoning for complex ones. In the xAI API, developers can fine-tune this behavior, optimizing for speed or depth.\n\nGrok 4 Fast is available now for all users. In `Fast` and `Auto` modes, you will see a significant improvement in search and information seeking queries. Additionally, difficult queries in `Auto` mode will use Grok 4 Fast, which will provide a much faster experience without loss of quality. For the first time, all users, including free users, will have access to our latest model without restrictions, marking a step toward democratizing advanced AI.\n\n## [Grok 4 Fast on OpenRouter, Vercel AI Gateway, and the xAI API](https://x.ai/news/grok-4-fast#grok-4-fast-on-openrouter-vercel-ai-gateway-and-the-xai-api)\n\nWe're rolling out Grok 4 Fast as two models: `grok-4-fast-reasoning` and `grok-4-fast-non-reasoning`, each with a 2M token context window. This allows developers to tune the amount of test-time compute applied to their use cases.\n\n`grok-4-fast-reasoning` and `grok-4-fast-non-reasoning` are generally available via the [xAI API](https://x.ai/api) according to the following pricing:\n\n| Token Type          | <128k tokens | ≥128k tokens |\n| ------------------- | ------------ | ------------ |\n| Input tokens        | $0.20 / 1M   | $0.40 / 1M   |\n| Output tokens       | $0.50 / 1M   | $1.00 / 1M   |\n| Cached input tokens | $0.05 / 1M   |              |\n\n## [What's Next](https://x.ai/news/grok-4-fast#whats-next)\n\nWe will continuously ship model improvements to Grok 4 Fast based on your feedback on [x.com](http://x.com). Stay tuned for further integrations, including enhanced multimodal capabilities and agentic features.\n\nRead the Grok 4 Fast model card [here](https://data.x.ai/2025-09-19-grok-4-fast-model-card.pdf).\n\nThat's all for now - so long, and thanks for all the fish!\n",
  "dateCreated": "2026-10-19T01:26:53.998049939+00:00",
  "date_created": "2025-09-19T00:00:00+00:00",
  "date_modified": null,
  "date_published": "2025-09-19T00:00:00+00:00",
  "description": "Pushing the Frontier of Cost-Efficient Intelligence",
  "external_url": "https://x.ai/news/grok-4-fast",
  "git": {
    "commit": "ebb0941c5c06a2ab2377342d607535262a3e969e",
    "path": "items/x-ai/2025/09/2025-09-19-grok-4-fast.md",
    "permalink": "https://github.com/aymericbeaumet/aggr-instance/blob/ebb0941c5c06a2ab2377342d607535262a3e969e/items/x-ai/2025/09/2025-09-19-grok-4-fast.md"
  },
  "id": "urn:aggr:item:5b5c1ea8190d757e8738ee6cefda0154d9d92327",
  "isBasedOn": {
    "@id": "https://x.ai/news/grok-4-fast"
  },
  "mainEntity": {
    "@id": "https://x.ai/news/grok-4-fast",
    "@type": "CreativeWork",
    "archivedAt": {
      "@id": "https://aggr.aymericbeaumet.com/items/x-ai/2025-09-19-grok-4-fast/#webpage"
    },
    "datePublished": "2025-09-19T00:00:00+00:00",
    "headline": "Grok 4 Fast",
    "image": {
      "@type": "ImageObject",
      "caption": null,
      "height": 90,
      "url": "https://media.x.ai/v1/website/grok-4-fast-8c8d3e84.webp",
      "width": 160
    },
    "timeRequired": "PT4M",
    "url": "https://x.ai/news/grok-4-fast",
    "wordCount": 872
  },
  "name": "Archived snapshot: Grok 4 Fast",
  "preview": {
    "alt": null,
    "color": null,
    "height": 90,
    "placeholder": {
      "data_url": "data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAQAAAC1HAwCAAAAC0lEQVR42mNk+A8AAQUBAScY42YAAAAASUVORK5CYII=",
      "hash": ""
    },
    "url": "https://media.x.ai/v1/website/grok-4-fast-8c8d3e84.webp",
    "width": 160
  },
  "reading_minutes": 4,
  "replicated_at": null,
  "source": "x-ai",
  "source_name": "News: Research, Product & Company Updates | SpaceXSI",
  "summary": "Pushing the Frontier of Cost-Efficient Intelligence",
  "tags": [],
  "temporalCoverage": "2026-10-19T01:26:53.998049939+00:00",
  "title": "Grok 4 Fast",
  "url": "https://aggr.aymericbeaumet.com/items/x-ai/2025-09-19-grok-4-fast/",
  "word_count": 872
}