Claude, GPT, Gemini, Grok, Mulu Agent 1, and every other model worth using, all inside Mulu Code. Switch between them freely, compare outputs in a glance, and pick the right brain for every job.
Switch models with a single keystroke. Mulu Code handles provider differences under the hood. You just pick the model best suited to the task in front of you.
| Model | Provider | Strengths | Context | Pricing |
|---|---|---|---|---|
| Mulu Agent 1 | Mulu | Reasoning, coding, agentic tool use | 200K | $2.50 / $8.00 per 1M |
| Claude Sonnet 4.6 | Anthropic | Coding, reasoning, extended thinking | 1M | Competitive |
| Claude Opus 4.6 | Anthropic | Deep reasoning, complex systems | 1M | Competitive |
| Claude Haiku 4.5 | Anthropic | Fast, cheap, light coding tasks | 200K | Competitive |
| GPT-5.4 | OpenAI | Reasoning, broad general purpose | 1M | Competitive |
| GPT-5.3 Codex | OpenAI | Coding specialist, fast iteration | 400K | Competitive |
| Gemini 3.1 Pro | Deep Think, long-context, UI design | 1M | Competitive | |
| Gemini 3 Flash | Fast, cheap, 1M context | 1M | Competitive | |
| Grok 4.2 | xAI | Reasoning toggle, 2M context | 2M | Competitive |
| Grok 4.2 Agents | xAI | Multi-agent orchestration variant | 2M | Competitive |
| Kimi K2.5 | Moonshot | 76.8% SWE-bench, 1T MoE | 256K | Competitive |
| Qwen 3.6 Plus | Alibaba | 76.4% SWE-bench, reasoning support | 1M | Competitive |
| MiniMax M2.7 | MiniMax | Strong reasoning, broad tasks | 200K | Competitive |
| MiMo v2 Pro | Xiaomi | 1T+ MoE (42B active), near-Opus coding | 1M | Competitive |
| MiMo v2 Flash | Xiaomi | Fast iteration, long context | 256K | Competitive |
| GLM-5 | Zhipu | Balanced reasoning and coding | 200K | Competitive |
| NVIDIA Nemotron 3 Super | NVIDIA | 120B MoE (12B active), efficient | 1M | Competitive |
Turn on Auto and Mulu routes each message to the best model for the task: Gemini 3.1 Pro for UI design, Claude Opus 4.6 for complex systems, GPT-5.3 Codex for fast code iteration, Mulu Agent 1 for agentic tool-calling work. Latency, quality, and cost are balanced automatically.
Override at any time. Force a specific model for a single message with a dropdown, or lock a whole conversation to a preferred model. Auto is a sensible default, not a cage.
Power users see which model handled each reply and can swap retroactively. "Redo this with Opus" is a single click on a message.
Ask the same question of three, four, or five models at once and read their responses side by side. Useful for important architectural decisions, unfamiliar APIs, and debugging sessions where you want a second opinion without switching tabs.
Each response is tagged with the model that produced it, and tokens used. You can cherry-pick a single reply to continue the thread with, or merge the best elements of several into a final answer. The orchestrator handles the prompt plumbing for you.
For long-term decisions, save comparison sessions to the project memory. Six months later, when you wonder "which model got this right?", the record is still there.
Mulu Agent 1 is our in-house reasoning and coding model, tuned specifically for agentic work inside Mulu Code. It's strong at multi-step tool use, code editing, and tasks that require following a plan across many turns without losing thread.
Priced at $2.50 per million input tokens and $8.00 per million output tokens. On the Power plan it's available at a dedicated per-month token pool for power users who want to run heavy workloads on one model.
You don't need to know what's inside. Just know it's tuned for the work you're doing here, and it keeps getting better.
Every model runs through Mulu's managed infrastructure by default. No separate signups, no juggling five provider accounts. If you'd rather use your own API keys from Anthropic, OpenAI, Google, or xAI, paste them into settings and the app routes your requests directly.
BYO keys bypass Mulu's billing entirely for those models. Your keys are stored in the OS keychain (Keychain, DPAPI, libsecret), never written to a plaintext config file, never synced to our servers.
You can mix and match: BYO keys for one provider, Mulu's managed billing for another. Teams can standardize on managed billing with central invoicing while letting individual contributors use personal keys when they prefer.
Download Mulu Code free and try any of the 17 models. No separate accounts, no per-provider setup.