Skip to content
Sign in

SpaceXAI's frontier model for coding, agentic tasks, and knowledge work.

Context window500K
ProviderGrok
Released2026/08/24

Playground

Pricing

The same model is available through multiple service channels — choose based on latency, reliability and cost.

Prices in $ / 1M tokens
To pick a channel, add a provider field to the request body, for example "provider": { "channel": "direct" }. Valid values are direct / stable / economical; omit it to use the default channel.

Direct

Direct upstream connection — best when you need native behavior and the full context window.

InputOutputCache read
2.00/M6.00/M0.50/M

Overview

Input
Text Image
Output
Text

Grok 4.6 API: A Long-Horizon Agentic Model in the Top Tier of Knowledge Work

Grok 4.6 is the flagship model released by xAI on August 12, 2026. Building on Grok 4.5, it significantly strengthens long-horizon agentic and interactive visual engineering capabilities. It scores 61 on the Artificial Analysis Intelligence Index, tying with GPT-5.6 Sol and coming within 1 point of Fable 5 Max — while remaining in Grok's customary low price tier.

Grok 4.6 is not a brand-new foundation model, but an incremental training run on Grok 4.5's ~1.5 trillion parameter base using curated engineering data and improved reinforcement learning. The most intuitive change in this upgrade: the model begins to proactively self-check in long tasks — verifying its own output before continuing, rather than running through once and calling it done.

SeaWhale AI offers Grok 4.6 through an OpenAI-compatible interface, supporting tool calling, streaming output, and multimodal input.

Get API Key · Model ID: grok-4.6


Why Choose Grok 4.6

  • #1 Across Knowledge Work — Champion in three benchmarks: GDPval-AA v2 (1753 points), AA-Briefcase, and Harvey Legal Evaluation
  • Intelligence Index Ties GPT-5.6 Sol — AA Intelligence Index of 61, up 5 points from Grok 4.5
  • Major Leap in Coding — DeepSWE v1.1 up from 54% to 65.9%, CursorBench 3.2 at 69.9%
  • Self-Checking in Long Tasks — Actively verifies intermediate outputs, reducing silent drift in long-chain tasks
  • 500K Token Context — Fits large codebases and complete datasets in one go

Core Capabilities

01 Long-Horizon Agentic Execution

Grok 4.6 continues the 4.5 optimization direction of "working for extended periods on real projects," and further focuses on persistence in multi-step tasks: researching a topic, analyzing a batch of materials, making changes across a codebase, turning an idea into a usable application — all while maintaining goal consistency across a large number of steps.

  • Goal consistency in multi-step tasks
  • Proactive self-checking and result verification in long chains
  • Sustained engineering changes across files and modules

02 Knowledge Work & Professional Domains

Currently in the top tier on real professional task benchmarks: first place in all three — GDPval-AA v2, AA-Briefcase, and Harvey Legal Evaluation. Document-intensive work in law, consulting, and analysis is where its advantage is most concentrated.

  • Legal document analysis and compliance review
  • Business analysis and professional report generation
  • Comprehensive research on large-scale materials

03 Interactive & Visual Engineering

This generation has clearly strengthened the ability to go "from rough product concept to usable prototype," with higher first-pass output quality on visual and interactive projects and less rework.

  • One-shot realization from product concept to functional prototype
  • Web application and interactive interface development
  • Significantly improved first-iteration quality

04 Major Leap in Coding

The incremental training used curated engineering data, pushing coding benchmarks up across the board: DeepSWE v1.1 65.9% (4.5 was 54%), CursorBench 3.2 69.9%, FrontierCode v1.1 61.3%. Coverage extends from routine development to kernel optimization and CAD environments.

  • Bug fixing and feature implementation in real codebases
  • Low-level engineering tasks such as kernel optimization
  • Vertical environments like web development and CAD

Best Use Cases

Scenario Description
Knowledge work automation Current #1 on real professional task benchmarks such as GDPval-AA v2
Legal & compliance analysis Harvey Legal Evaluation champion, clear advantage in document-intensive tasks
Long-horizon coding agent Codebase-level changes across many steps, self-checking reduces drift
Rapid prototyping From product concept directly to interactive functional prototype
Deep research 500K token context combined with multi-step research workflows
Cost-sensitive production deployment Frontier capabilities at Grok's price tier

Grok 4.6 vs. Grok 4.5 and GPT-5.6 Sol

Capability Grok 4.6 Grok 4.5 GPT-5.6 Sol
Model ID grok-4.6 grok-4.5 gpt-5.6-sol
Vendor xAI xAI OpenAI
AA Intelligence Index 61 56 61
DeepSWE v1.1 65.9% 54% Higher (leads in this metric)
GDPval-AA v2 1753 (#1) Behind Grok 4.6
Context Window 500K tokens 500K tokens 1M+ tokens
Input Modalities Text + Image Text + Image Text + Image

For specific billing, refer to the real-time price card at the top of the page.


FAQ

When was Grok 4.6 released? Released on August 12, 2026.

What are the upgrades compared to Grok 4.5? Three areas: first, the Intelligence Index went from 56 to 61, with coding benchmarks rising across the board (DeepSWE from 54% to 65.9%); second, it took first place in three knowledge work benchmarks; third, proactive self-checking behavior emerged in long tasks — verifying before proceeding. It is an incremental training run on the 4.5 base, not a new foundation model.

Any areas where it lags behind? Yes. It loses to GPT-5.6 Sol on DeepSWE v1.1 and Terminal-Bench 3.0 — Sol still holds a slight edge on pure code execution benchmarks, but Grok 4.6's price tier is much lower.

How large is the context window? 500K tokens, supports text and image input, with text output.

What is it best suited for? Two types of tasks best highlight the difference: document-intensive professional knowledge work (law, analysis, research), and agentic workflows that need to span many steps. Short one-off Q&A won't show the difference from the previous generation.

How do I use reasoning mode? Grok 4.6 provides an explicit reasoning mode. Enabling it for complex tasks yields a more complete reasoning chain; leaving it off for simple tasks saves tokens.


Why Use SeaWhale AI for Grok 4.6 API

  • No overseas account needed — Direct connection in China, no need to set up a proxy
  • OpenAI-compatible interface — Connect with just a two-line change to existing code
  • Cross-vendor comparison — Compare prices and performance directly with Claude, GPT, and Gemini under the same account
  • Unified billing — Centrally manage usage and costs across multiple models

API

API integration

Model IDUse this value as the model in inference requests
grok-4.6
API KeyBearer token used to authenticate inference requests
Base URLOpenAI compatible · /chat/completions
OpenAIhttps://api.seawhaleai.com/v2/
provider OptionalSelects a service channel; omit it and the system picks the default
"provider": { "channel": "direct" }

grok-4.6 usage examples

SeaWhale AI is compatible with the OpenAI API protocol, so you can call it with the OpenAI SDK or plain HTTP requests. Streaming is enabled by default.

About the provider parameter (optional, a SeaWhale AI extension): most models are served over several channels that differ slightly in price and reliability. Add a provider field to the request body to pick one; omit it and the system selects the default channel — normal calls are unaffected.

provider is not part of the official OpenAI protocol — it is a SeaWhale AI extension that only takes effect on this platform. The OpenAI SDK allows custom fields like this to pass through; see the examples below.

ValueChannelBest for
directDirectThe official upstream link, for native behavior and the full context window
stablePreferredBalanced availability and speed — a good default for production traffic
economicalEconomyCost first, well suited to batch processing and price-sensitive workloads

Available channels and their prices are listed under "Pricing" above (channels vary by model). Additional notes:

  • Syntax: "provider": { "channel": "direct" }.
  • If the requested channel is not enabled for that model, the request automatically falls back to the default channel and returns normally — it does not error.
  • With the official SDKs: in Python pass it via extra_body; in Node.js put it directly on the request object and it passes through. In TypeScript projects, add a // @ts-expect-error line to skip the type check.
js
curl https://api.seawhaleai.com/v2/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer <API_KEY>" \
  -d '{
    "model": "grok-4.6",
    "messages": [
      {"role": "system", "content": "You are a helpful assistant."},
      {"role": "user", "content": "Hello!"}
    ],
    "provider": { "channel": "direct" },
    "stream": true
  }'
# provider is optional — remove this line to use the default channel
js
from openai import OpenAI

client = OpenAI(
    base_url="https://api.seawhaleai.com/v2",
    api_key="<API_KEY>",
)

stream = client.chat.completions.create(
    model="grok-4.6",
    messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "Hello!"},
    ],
    stream=True,
    # Optional: pick a service channel; omit to use the default
    extra_body={"provider": {"channel": "direct"}},
)

for chunk in stream:
    if chunk.choices[0].delta.content:
        print(chunk.choices[0].delta.content, end="", flush=True)
js
import OpenAI from 'openai'

const client = new OpenAI({
  baseURL: 'https://api.seawhaleai.com/v2',
  apiKey: '<API_KEY>',
})

const stream = await client.chat.completions.create({
  model: 'grok-4.6',
  messages: [
    { role: 'system', content: 'You are a helpful assistant.' },
    { role: 'user', content: 'Hello!' },
  ],
  stream: true,
  // Optional: pick a service channel; omit to use the default
  // @ts-expect-error provider is a SeaWhale AI extension, not in the OpenAI SDK types
  provider: { channel: 'direct' },
})

for await (const chunk of stream) {
  process.stdout.write(chunk.choices[0]?.delta?.content ?? '')
}