xAI's latest model, Grok 4.3, is now available on CodeCloud via OpenRouter. It joins our growing lineup of reasoning-capable models that can run coding agents in the cloud.

Grok 4.3 brings a few things that make it interesting for automated coding workflows: a 2 million token context window, native tool calling, and pricing that undercuts the frontier models while delivering competitive reasoning performance.

Why Grok 4.3 matters for coding agents

Model selection for coding agents isn't just about which one scores highest on benchmarks. The practical constraints matter more: how much context can it hold, does it reliably call tools, and what's the cost per run.

Grok 4.3 checks some important boxes:

2M token context – The agent can read your entire codebase, not just the files that fit in a smaller window. No chunking, no summarization, no missing context when it's working on a feature that touches 50 files across multiple directories
Native tool calling – Required for agents that need to run commands, read files, and interact with APIs. Grok 4.3 supports structured outputs and function calling out of the box
Reasoning capability – Complex coding tasks require multi-step thinking. The model can plan, execute, observe results, and adjust its approach without hand-holding
Cost-effective – At $1.25/M input tokens and $2.50/M output tokens via OpenRouter, it's priced below frontier models while still delivering strong performance

Using Grok 4.3 with CodeCloud

Set model to x-ai/grok-4.3 in your API request. The model is available through OpenRouter, which CodeCloud integrates with directly.

curl -X POST https://codecloud.dev/api/v1/agents \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "repo": "your-org/your-repo",
    "prompt": "Refactor the authentication module to use JWTs instead of session cookies. Update all related tests.",
    "model": "x-ai/grok-4.3",
    "auto_create_pr": true
  }'

The large context window makes Grok 4.3 especially useful for tasks that require understanding the full codebase — architectural refactors, cross-cutting changes, and features that touch many files.

When to use Grok 4.3 vs other models

We now support a wide range of models, each with different trade-offs. Here's how to think about model selection:

Grok 4.3 – Best for large codebase analysis, cross-cutting refactors, and tasks where you need the agent to see everything. The 2M context window eliminates the need to guess which files are relevant
Claude Opus 4.7 – Best for complex reasoning, nuanced architectural decisions, and tasks where quality matters more than cost. The frontier model for difficult problems
Claude Sonnet 4.6 – The balanced choice for most coding tasks. Good reasoning, reasonable cost, fast enough for interactive use
Claude Haiku 4.5 – Best for high-volume, simple tasks. Dependency updates, typo fixes, small bug fixes. When you're running 50+ agents a day

The right model depends on the task. A typo fix doesn't need Opus. A major architectural refactor probably shouldn't use Haiku. Grok 4.3 fills the niche of "I need the agent to understand my entire codebase at once."

OpenRouter integration

Grok 4.3 is available through OpenRouter, which means you bring your own OpenRouter API key and pay OpenRouter directly. CodeCloud handles the agent orchestration; OpenRouter handles the model routing.

This architecture keeps costs transparent — you pay the model provider's rate with no markup from us. It also means you get access to new models as they're released on OpenRouter without waiting for us to add support.

The model landscape for agents

The AI model market has evolved quickly. Two years ago, there was one frontier model and everything else was significantly worse. Now we have multiple providers (Anthropic, OpenAI, Google, xAI, Meta, Mistral) all producing models that can handle coding tasks.

This competition is good for developers building automated coding workflows:

Choice – Pick the right model for the task instead of being locked into one provider
Price pressure – Costs have dropped significantly as providers compete
Specialization – Some models optimize for speed, others for reasoning depth, others for context size
Reliability – If one provider has an outage, you can fall back to another

CodeCloud's job is to make all of these models work as coding agents. You pick the model, we handle the sandbox, the file system, the git operations, and the tool execution.

Getting started

If you already have an OpenRouter account, add your API key in the CodeCloud dashboard and you can start using Grok 4.3 immediately. If you're new to OpenRouter, sign up at openrouter.ai.

For a full list of supported models, see the documentation. Questions or feedback? Reach out at support@codecloud.dev.