Claude Code: Compatible Models & Providers

A comprehensive reference guide for developers using Claude Code with official and alternative AI providers. This guide covers the latest flagship models, pricing (Input/Output per 1M tokens), configuration for Anthropic, Alibaba Qwen, DeepSeek, MiniMax, Moonshot AI (Kimi), Zhipu GLM, Xiaomi MiMo, StepFun, and multi-provider gateways.

Compatible Providers

| Provider | Type | Notable Models | Coding Plan | | :--- | :--- | :--- | :--- | | Anthropic | Official | Claude Opus 4.7, Sonnet 4.6 | Max subscription | | Alibaba (DashScope) | Direct | Qwen 3.6-Plus, Qwen 3.5-Coder | — | | DeepSeek | Direct | DeepSeek V3.2, DeepSeek-R2 | — | | MiniMax | Direct | MiniMax-M2.7, M2.7-Highspeed | Token Plan (from $10/mo) | | Moonshot (Kimi) | Direct | Kimi K2.5, K2.6 (beta) | — | | Zhipu (Z.ai) | Direct | GLM-5.1, GLM-5, GLM-4.7 | Coding Plan (from $18/mo) | | Xiaomi (MiMo) | Direct | MiMo-V2-Pro, V2-Omni, V2-Flash | Token Plan (from ~$6/mo) | | StepFun | Direct | Step-3.5-Flash, Step-3 | Step Plan (from $6.99/mo) | | SiliconFlow | Gateway | Hosted DeepSeek, GLM, Qwen, Kimi | — | | OpenRouter | Gateway | 200+ models from all providers | — | | Ollama | Local / Cloud | Open-source models, free | Free (local) |

---

Provider Setup Guides

Anthropic (Official)

Claude Code officially supports Claude Opus 4.7 and Sonnet 4.6, excelling at agentic coding tasks and tool use.

Installation:

npm install -g @anthropic-ai/claude-code

Authentication:

• OAuth: Run claude and follow the browser login.
• API Key: Set ANTHROPIC_API_KEY in your environment.

Official Documentation

---

Alibaba Cloud - DashScope (Qwen)

Qwen 3.6-Plus is the latest flagship (April 2026), with 1M context window and strong agentic coding capabilities.

Configuration: Edit ~/.claude/settings.json:

{
  "env": {
    "ANTHROPIC_BASE_URL": "https://dashscope-intl.aliyuncs.com/apps/anthropic",
    "ANTHROPIC_AUTH_TOKEN": "YOUR_DASHSCOPE_API_KEY",
    "ANTHROPIC_MODEL": "qwen3.6-plus"
  }
}

Pricing tiers (per 1M tokens): | Context Band | Input | Output | | :--- | :--- | :--- | | 0 - 256K | $0.28 | $1.65 | | 256K - 1M | $1.10 | $6.60 |

---

DeepSeek

DeepSeek V3.2 and R2 offer incredibly cost-effective coding and reasoning.

Configuration:

{
  "env": {
    "ANTHROPIC_BASE_URL": "https://api.deepseek.com/anthropic",
    "ANTHROPIC_AUTH_TOKEN": "YOUR_DEEPSEEK_API_KEY",
    "ANTHROPIC_MODEL": "deepseek-reasoner"
  }
}

Pricing (per 1M tokens): | Model | Input | Cache Hit | Output | | :--- | :--- | :--- | :--- | | DeepSeek V3.2 | $0.27 | $0.07 | $1.10 | | DeepSeek-R2 | $0.55 | $0.14 | $2.19 |

Note: deepseek-reasoner maps to the reasoning model with thinking processes.

---

MiniMax

MiniMax-M2.7 (March 2026) is open-source, 230B params MoE, with SOTA on real-world software engineering benchmarks.

Configuration:

{
  "env": {
    "ANTHROPIC_BASE_URL": "https://api.minimax.io/anthropic",
    "ANTHROPIC_AUTH_TOKEN": "YOUR_MINIMAX_API_KEY",
    "ANTHROPIC_MODEL": "minimax-m2.7"
  }
}

China endpoint: https://api.minimaxi.com/anthropic

Pricing (per 1M tokens): | Model | Input | Cache Read | Output | | :--- | :--- | :--- | :--- | | MiniMax-M2.7 | $0.30 | $0.059 | $1.20 | | M2.7-Highspeed | $0.60 | — | $2.40 |

---

Moonshot AI - Kimi

Kimi K2.5 is a 1T-parameter MoE model optimized for long context. K2.6 is in beta (April 2026).

Configuration:

{
  "env": {
    "ANTHROPIC_BASE_URL": "https://api.moonshot.ai/anthropic",
    "ANTHROPIC_AUTH_TOKEN": "YOUR_MOONSHOT_API_KEY",
    "ANTHROPIC_MODEL": "kimi-k2.5"
  }
}

Note: K2.6 is currently beta-only via Kimi CLI/IDE integrations. Public API expected May 2026.

---

Zhipu (Z.ai) - GLM

GLM-5.1 (April 2026) is a 744B MoE model capable of 8+ hour autonomous tasks, MIT-licensed.

Configuration:

{
  "env": {
    "ANTHROPIC_BASE_URL": "https://api.z.ai/api/anthropic",
    "ANTHROPIC_AUTH_TOKEN": "YOUR_ZAI_API_KEY",
    "ANTHROPIC_MODEL": "glm-5.1"
  }
}

Pricing (per 1M tokens): | Model | Input | Output | | :--- | :--- | :--- | | GLM-5.1 | $0.95 - $1.40 | $3.15 - $4.40 | | GLM-5 | $1.00 | $3.20 | | GLM-4.7 | $0.60 | $2.20 |

Note: GLM-5.1 consumes 3x quota during peak hours, 2x off-peak on Coding Plan.

---

Xiaomi - MiMo

MiMo-V2-Pro ranks top-5 globally on OpenRouter Arena. 1M context, MoE architecture.

Configuration:

{
  "env": {
    "ANTHROPIC_BASE_URL": "https://api.xiaomimimo.com",
    "ANTHROPIC_AUTH_TOKEN": "YOUR_MIMO_API_KEY",
    "ANTHROPIC_MODEL": "mimo-v2-pro"
  }
}

Pricing (per 1M tokens): | Model | Context | Input | Output | Cache Read | | :--- | :--- | :--- | :--- | :--- | | MiMo-V2-Pro | ≤256K | $1.00 | $3.00 | $0.20 | | MiMo-V2-Pro | 256K-1M | $2.00 | $6.00 | $0.40 | | MiMo-V2-Omni | 256K | $0.40 | $2.00 | $0.08 | | MiMo-V2-Flash | 256K | $0.09 | $0.29 | $0.01 |

Supported integrations: Claude Code, OpenCode, OpenClaw, Cline, Kilo Code, Roo Code, Codex, Cherry Studio, Zed.

---

StepFun

Step-3.5-Flash is a powerful reasoning model at the lowest price point among major providers.

Configuration (Pay-as-you-go):

{
  "env": {
    "ANTHROPIC_AUTH_TOKEN": "YOUR_STEPFUN_API_KEY",
    "ANTHROPIC_BASE_URL": "https://api.stepfun.ai/",
    "ANTHROPIC_MODEL": "step-3.5-flash"
  }
}

Configuration (Step Plan subscription):

{
  "env": {
    "ANTHROPIC_API_KEY": "YOUR_STEPFUN_API_KEY",
    "ANTHROPIC_BASE_URL": "https://api.stepfun.ai/step_plan",
    "ANTHROPIC_MODEL": "step-3.5-flash"
  }
}

China endpoints: Replace .ai with .com (e.g., https://api.stepfun.com/).

Pricing (per 1M tokens): | Model | Input | Output | Cache Read | | :--- | :--- | :--- | :--- | | Step-3.5-Flash | $0.10 | $0.30 | $0.02 | | Step-3 | ~$0.21 | ~$0.55 - $1.10 | ~$0.04 | | Step-2-Mini | ~$0.14 | ~$0.28 | ~$0.03 |

Supported integrations: Claude Code, OpenCode, OpenClaw, Cline, Kilo Code, Roo Code, Trae, Cursor, Zed, Cherry Studio, Goose.

---

SiliconFlow (Multi-Provider)

Hosted access to DeepSeek, GLM, Qwen, Kimi, and more via Anthropic-compatible API.

Configuration:

{
  "env": {
    "ANTHROPIC_BASE_URL": "https://api.siliconflow.com/",
    "ANTHROPIC_API_KEY": "YOUR_SILICONFLOW_KEY",
    "ANTHROPIC_MODEL": "deepseek-ai/DeepSeek-V3"
  }
}

Note: Currently does not support "thinking" model variants — only non-thinking.

---

Ollama (Local / Cloud)

Run models locally for free since January 2026 (v0.14.0+). Also supports cloud models via ollama.com.

Configuration:

{
  "env": {
    "ANTHROPIC_AUTH_TOKEN": "ollama",
    "ANTHROPIC_BASE_URL": "http://localhost:11434",
    "ANTHROPIC_API_KEY": ""
  }
}

Quick start: ollama launch claude --config provides guided setup. Requires models with 32K+ context.

Local models: qwen3-coder, glm-4.7, minimax-m2.1, and more. Cloud models: glm-5.1:cloud, kimi-k2.5:cloud, qwen3.5:cloud via ollama.com.

---

OpenRouter (Multi-Provider)

Best for comparing models or failover. Supports all providers via one token, including Fast Mode (2.5x faster output at premium pricing).

Basic Setup:

export ANTHROPIC_BASE_URL="https://openrouter.ai/api"
export ANTHROPIC_AUTH_TOKEN="YOUR_OPENROUTER_KEY"

Model environment variables:

export ANTHROPIC_DEFAULT_OPUS_MODEL="anthropic/claude-opus-4.7"
export ANTHROPIC_DEFAULT_SONNET_MODEL="anthropic/claude-sonnet-4.6"

---

Coding Plans & Subscriptions

Several providers offer subscription plans with flat-rate usage for Claude Code, typically measured in requests within a 5-hour rolling window.

StepFun — Step Plan

| Plan | Price | Prompts / 5 hrs | Est. Model Calls | | :--- | :--- | :--- | :--- | | Flash Mini | $6.99/mo | 100 | ~1,500 | | Flash Plus | $9.99/mo | 400 | ~6,000 | | Flash Pro | $29/mo | 1,500 | ~22,500 | | Flash Max | $99/mo | 5,000 | ~75,000 |

All plans include step-3.5-flash, multi-device login, and concurrent agent execution. Plus and above get priority API access and support.

Step Plan pricing →

---

Xiaomi MiMo — Token Plan

| Plan | Price | Credits | Est. Complex Tasks | | :--- | :--- | :--- | :--- | | Lite | ~$6/mo (39 RMB) | 60M | ~120 | | Standard | ~$16/mo (99 RMB) | 200M | ~400 | | Pro | ~$50/mo (329 RMB) | 700M | ~1,400 | | Max | ~$100/mo (659 RMB) | 1.6B | ~3,200 |

Credit multipliers: MiMo-V2-Omni 1x, MiMo-V2-Pro (≤256K) 2x, MiMo-V2-Pro (256K-1M) 4x. First purchase gets 12% off. No 5-hour limit.

Token Plan pricing →

---

MiniMax — Token Plan

| Plan | Price | Requests / 5 hrs | | :--- | :--- | :--- | | Starter | $10/mo | 1,500 | | Plus | $20/mo | 4,500 | | Max | $50/mo | 15,000 | | Max-Highspeed | $80/mo | 15,000 | | Ultra-Highspeed | $150/mo | 30,000 |

All tiers include MiniMax-M2.7. Highspeed plans use the faster M2.7-Highspeed variant. Yearly plans available: Starter $100/yr, Plus $200/yr, Max $500/yr.

Token Plan pricing →

---

Zhipu (Z.ai) — Coding Plan

| Plan | Price | Quota | | :--- | :--- | :--- | | Lite | $18/mo | 3x Claude Pro usage | | Pro | $72/mo | 5x Lite Plan usage | | Max | $160/mo | 4x Pro Plan usage |

GLM-5.1 consumes 3x quota during peak hours, 2x off-peak. Pro and above include Vision and Web Search.

Coding Plan pricing →

---

Alibaba Cloud — Model Studio Coding Plan

| Plan | Price | Quota | Status | | :--- | :--- | :--- | :--- | | Lite | ~$10/mo | Lower quotas | Discontinued for new purchases (Mar 2026) | | Pro | $50/mo | 6,000 req/5hrs, 90K req/mo | Available |

Pro plan includes Qwen 3.5-Plus, Kimi K2.5, GLM-5, MiniMax-M2.5, and Qwen 3-Coder variants. Each query costs 5-30+ requests depending on complexity.

Coding Plan pricing →

---

Moonshot (Kimi) — Kimi Code

| Plan | Price | Quota | | :--- | :--- | :--- | | Free | $0 | Basic chat, daily/weekly quotas | | Moderato | ~$19/mo | 2,048 Kimi Code requests/week | | Allegretto | ~$49/mo | Higher quotas | | Vivace | ~$129/mo | Highest quotas, premium features |

Plans include CLI and developer tools, priority compute, and 5-hour rolling token quotas.

Kimi Code →

---

Ollama — Cloud Plans

| Plan | Price | Cloud Usage | Concurrent Cloud Models | | :--- | :--- | :--- | :--- | | Free | $0 | Light | 1 | | Pro | $20/mo ($200/yr) | 50x Free | 3 | | Max | $100/mo | 5x Pro | 10 |

Local model execution is always unlimited and free on all plans. Cloud usage resets every 5 hours (session) and weekly.

Ollama Cloud pricing →

---

Quick Tips

1. Thinking Mode: For complex refactors, use models like deepseek-reasoner, kimi-k2-thinking, or mimo-v2-flash-thinking.
2. Context Caching: DeepSeek, Anthropic, and MiMo support caching — reduces input costs by up to 90%.
3. Regional Endpoints: In China, use open.bigmodel.cn for GLM, dashscope.aliyuncs.com for Qwen, api.stepfun.com for StepFun, or api.minimaxi.com for MiniMax for lower latency.
4. Step Plan API: Use https://api.stepfun.ai/step_plan (not the standard endpoint) for Step Plan subscriptions.
5. Ollama Local: Run ollama launch claude --config for guided zero-cost local setup with open-source models.
6. Free Tiers: Step-3.5-Flash and MiMo-V2-Flash are available free (rate-limited) via OpenRouter.