For OpenCode users · hosted frontier open models
OpenCode, now with GLM-5, Kimi K3 & DeepSeek V4
Point OpenCode at one OpenAI-compatible API and get 23 coding models — flagship GLM-5, Kimi K3, DeepSeek V4 Pro, Qwen3 Coder and more. Coding inference from $0.04 / 1M tokens, plans from $5/mo, no rate limits.
OpenCode is a fast, open-source terminal coding agent. The friction is the model layer: provider free tiers throttle you, and every model family means another API key and another base URL to manage.
AINative removes that friction. One OpenAI-compatible endpoint serves 23 coding models — flagship GLM-5, Kimi K2 through K3, DeepSeek V4 Pro, MiniMax M2.5/M2.7, Qwen3 Coder, plus Claude Sonnet 4.5 / Opus 4 and GPT-5 on paid tiers — with native tool calling, no rate-limit tiers, and transparent per-token pricing (from $0.04 / 1M tokens). Change your base URL, drop in one key, and switch models in OpenCode without re-plumbing.
OpenCode on AINative vs. a single provider
| Feature | OpenCode + AINative | Single provider |
|---|---|---|
| Models | 23 coding models — GLM-5, Kimi K3, DeepSeek V4, Qwen3 Coder, Claude, GPT-5 | Whatever provider you wire up |
| Rate limits | None — pay-as-you-go throughput | Provider free tiers throttle hard |
| API surface | One OpenAI-compatible endpoint | Different base URL per provider |
| Cheapest inference | Qwen3 Coder 30B at $0.04 / 1M tokens | Rarely this cheap |
| Entry-plan coding models | 5 on Hobbyist ($5/mo) — incl. GPT OSS 120B and Qwen3 Coder | Usually paywalled |
| Setup | Change base_url + one API key | Manage a key per provider |
| Agent memory | ZeroDB persistent memory built in | Bring your own |
| Tool calling | Native function calling on all agent models | Model-dependent |
| Best for | OpenCode users who want frontier models, cheap | Single-provider setups |
Why route OpenCode through AINative
Frontier + open models
23 coding models — flagship GLM-5, Kimi K3, DeepSeek V4 Pro, Qwen3 Coder, plus Claude Sonnet 4.5 / Opus 4 and GPT-5 on paid tiers — all behind one endpoint.
No rate limits
Skip the free-tier throttling. Pay-as-you-go throughput means your coding agent keeps working when a provider free tier would have cut you off.
One key, one endpoint
OpenAI-compatible API. Change the base URL and drop in a single key — no juggling a separate credential per model family.
Persistent agent memory
ZeroDB gives your agent vector + graph memory that survives sessions, so context and decisions carry forward across runs.
Coding model pricing
23 coding models across all plans. All prices per 1M tokens. Verified serving 2026-07-28.
Hobbyist — $5/mo
5 coding models · 7-day free trial| Model | Input / 1M | Output / 1M | Context |
|---|---|---|---|
| Gemma 4 31B | $0.99 | $1.49 | 64K |
| GLM-4.7 (Preview) | $1.25 | $2.75 | 8K |
| GPT OSS 120B | $0.42 | $0.90 | 64K |
| Qwen3 Coder 30B | $0.04 | $0.04 | 256K |
| Qwen3 Coder Next 80B | $0.06 | $0.06 | 256K |
Pro — $49/mo
Frontier + newest open models| Model | Input / 1M | Output / 1M | Context |
|---|---|---|---|
| Qwen Coder 7B | $0.24 | $0.96 | 128K |
| Qwen3 Coder Flash | $0.15 | $0.60 | 128K |
| DeepSeek 4 Flash | $0.20 | $0.80 | 128K |
| Kimi K2 | $0.25 | $1.00 | 128K |
| Kimi K2.5 | $0.25 | $1.00 | 128K |
| Kimi K2.6 | $0.25 | $1.00 | 128K |
| Qwen Coder 32B | $0.28 | $1.10 | 128K |
| MiniMax M2.5 | $0.30 | $1.20 | 128K |
| MiniMax M2.7 | $0.60 | $2.40 | 128K |
| Qwen3.5 397B MoE | $0.30 | $1.20 | 128K |
| DeepSeek V4 Pro | $0.44 | $0.87 | 128K |
| GLM-5 | $0.60 | $2.20 | 128K |
| GLM-5.1 | $0.60 | $2.20 | 128K |
| GLM-5.2 | $0.60 | $2.20 | 128K |
| Kimi K3 | $0.60 | $2.50 | 256K |
| Claude Sonnet 4.5 | $6.00 | $30.00 | 195K |
Enterprise — $999/mo
| Model | Input / 1M | Output / 1M | Context |
|---|---|---|---|
| Claude Opus 4 | $15.00 | $75.00 | 195K |
Get your API key in 30 seconds
Get started with Cody CLI
No signup required. Install and start coding in 30 seconds.
Hobbyist plan ($5/mo, 7-day free trial) includes 5 coding models, persistent memory, and MCP support
Get product updates
New models, features, and coding tips. No spam.
Frequently Asked Questions
Can I use AINative models with OpenCode?
Yes. AINative exposes an OpenAI-compatible API, so any tool that accepts a custom OpenAI base URL and API key — including OpenCode and other terminal coding agents — works out of the box. Point OpenCode at the AINative endpoint and you get Kimi K3, GLM-5, DeepSeek V4, Qwen3 Coder and 140+ more models through a single key.
Which coding models does AINative host, and what do they cost?
AINative hosts 23 coding models. The Hobbyist plan ($5/mo, 7-day free trial) includes GPT OSS 120B, Gemma 4 31B and Qwen3 Coder (from $0.04/1M tokens). Pro ($49/mo) adds GLM-5/5.1/5.2, Kimi K2/K2.5/K2.6/K3, DeepSeek 4 Flash / V4 Pro, MiniMax M2.5/M2.7, Qwen3.5 397B MoE and Claude Sonnet 4.5. Enterprise adds Claude Opus 4. All prices are per 1M tokens with native tool calling and no per-model rate-limit tiers.
What does the entry plan for the coding models API cost?
The Hobbyist plan is $5/mo with a 7-day free trial. It includes API credits and access to the non-premium coding models. Sign up, grab your key, and point OpenCode (or any agent) at the AINative endpoint in seconds.
Why route OpenCode through AINative instead of a single provider?
Single providers throttle free tiers and force a separate API key and base URL for each model family. AINative gives you Kimi, GLM, DeepSeek, Qwen and 140+ models behind one key and one endpoint, with no rate-limit tiers and transparent pay-as-you-go pricing — so you can switch models in OpenCode without re-plumbing anything.
What is an open-source alternative to Claude Code?
OpenCode is a popular open-source, terminal-based coding agent. Paired with AINative-hosted frontier open models (Kimi K3, GLM-5, DeepSeek V4), it becomes a fully open alternative to Claude Code — no Anthropic subscription, no single-model lock-in, and no rate limits.
