Grok 4.6
grok-4-6grok-4.6- 📏 500K-token context window for long documents and extended agent runs.
- 🧠 Builds on Grok 4.5, focused on long-running agents and ambitious interactive work.
- 🔧 Adjustable reasoning effort: low, medium, high, and xhigh.
- 👁️ Multimodal input: text plus images, with vision-grounded reasoning.
- 🎯 Function calling, structured outputs, and integrated web search.
- 📚 Longer supplemental training run than 4.5, with improved optimizer and recipe.
- 💬 API model name is grok-4.6 via xAI's Responses or Chat Completions endpoints.
- 🏢 xAI's frontier text model for coding, agentic tasks, and knowledge work.
xAI is an American artificial intelligence company and wholly owned subsidiary of SpaceX. The company develops AI systems under the Grok brand, spanning language models, image generation, and video synthesis. xAI has quickly established itself as a multimodal AI lab with…
Explore 9 more models by XAI (company) →Grok 4.6 is xAI's frontier text-and-vision reasoning model, positioned for coding, agentic workflows, and knowledge work. It accepts mixed text and image input, supports function calling and structured outputs, can call an integrated web-search tool, and exposes an adjustable reasoning-effort setting spanning low, medium, high, and xhigh so developers can trade latency against depth. Its 500K-token context window suits multi-document synthesis and long tool-use sessions.
Compared with its direct predecessor Grok 4.5, xAI describes Grok 4.6 as building on that model with "a particular focus on long-running agents and more ambitious interactive and visual work." The lab reports that Grok 4.6 went through a longer supplemental training run than Grok 4.5, using curated model-generated data for reasoning and advanced technical concepts, high-quality engineering data, and an improved optimizer and training recipe. xAI also says it used Grok 4.5 itself to regenerate supervised fine-tuning trajectories across reasoning efforts, agent harnesses, and domains such as STEM, software engineering, and knowledge work, filtering problematic traces with model-based checks. In xAI's own testing, the model is described as especially strong at turning a broad product idea into a working first version, researching unfamiliar domains and refining results over several feedback rounds.
Within the wider Grok line, it succeeds Grok 4.3 and the Grok 4.20 generation, including the parallel-agent variant Grok 4.20 Multi-Agent, and complements agent-oriented Grok Build 0.1 plus xAI's separate visual models such as Grok Imagine 2.0. Note that xAI had not published a full benchmark table alongside the model's technical documentation at the time of writing, so no scores are quoted here.
This About section is AI-generated from public sources via VeniceStats + Venice inference, with no human editing. It may contain inaccuracies.
| Seller | Reputation↓ | Routing | Input $/M | Cached $/M | Output $/M | Categories | API |
|---|---|---|---|---|---|---|---|
| ▲ Apex Ant 0x73b4…e736 | 71.90 | #4 | $1.56 | $0.39 | $4.68 | chat,coding,reasoning,long-context,agents,premium,multimodal | openai-chat-completions |
| D5V1N2 0xd5e7…7be0 | 67.45 | #2 | $1.10 | $0.28 | $3.30 | x-search,web-search,chat,coding,reasoning,vision,large-context,grok,frontier,router,fallback | openai-chat-completions |
| Super Seeder 0xd19f…41f3 | 63.77 | #3 | $1.1237 | $0.2822 | $3.366 | chat,coding,reasoning,vision,multimodal,tools,long-context | openai-chat-completions |
| Edith AI 0xb269…b1a6 | 61.93 | #1 | $1.00 | $0.20 | $3.00 | chat,coding | openai-chat-completions |
| bartly.eth64.de 0x666e…4666 | 53.43 | gated | $0.3728 | $0.0932 | $1.1183 | chat,coding,reasoning,vision,multimodal,tools,long-context,web-search,function-calling | openai-chat-completions |
| Fire Ant 🔥🐜 0xbe05…bc5d | 50.00 | gated | $1.5068 | $0.3784 | $4.5137 | chat,coding,reasoning,tools | — |
| Hana Gateway ✅ 0x4ae1…117b | 45.41 | gated | $0.10 | $0.10 | $0.30 | chat,coding,reasoning,tools | openai-chat-completions |
"Best price" and the seller table are live AntSeed catalog data (advertised $/1M tokens — or $ per generated image for unit-billed image models — not settled amounts). Reputation = buyer trust score (0-100, the AntSeed SDK's own formula). "Routing" = the SDK's default buyer routing (what the VPR desktop app ships with): a trust ≥ 60 gate on the effective reputation, then cheapest-first among routable sellers; live failover state (per-peer cooldowns) is buyer-side runtime and not included. Model knowledge (TLDR, provider, About) via the VeniceStats enrichment layer. Advertised catalog, not the model used in any specific purchase. "Usage on AntSeed" counts only settlements whose buyers share the per-model split on-chain (metadata v2/v3, opt-in), so every usage figure is a lower bound.