Gwarden AIGwarden AI
25 models
A powerful model for code and agent tasks - via a simple API.
Get your key in Telegram, start with a free 7-day trial.
Everything for development
Agent-ready
Full tool-calling and structured output - for TUI agents and IDEs.
200K context
A huge window for large codebases and long sessions.
Max output
Reasoning + content as separate SSE chunks.
Launch in a minute
Get a key
Open the bot, subscribe to the channel, press «Get API».
Add config
Add baseURL and apiKey to your tool.
Done
Works with opencode, Claude Code and any OpenAI client.
# opencode - check the model
curl https://gwarden.su/v1/chat/completions \
-H "Authorization: Bearer GWAR-XXXX" \
-H "Content-Type: application/json" \
-d '{"model": "glm-5.3", "messages": [{"role": "user", "content": "Hello"}]}'
Subscriptions
Every 5 hours your balance refills to the plan limit. Start - free 7 days. Per 1M tokens: Input 0.15, Output 0.5, Cache 0.03. Top-up: 1$ = 10.0$.
Models & pricing
| Model | Input / 1M | Output / 1M | Cache / 1M | Quality |
|---|---|---|---|---|
| glm-5.3200k / 128k out | $0.15 | $0.50 | $0.03 | 59.5% |
| claude-fable-5200k / 128k out | $3.50 | $13.00 | $4.00 | 62.1% |
| claude-fable-5-1200k / 128k out | $3.50 | $13.00 | $4.00 | 65.7% |
| claude-opus-5200k / 64k out | $2.00 | $10.00 | $2.00 | 63.1% |
| claude-opus-4-8200k / 64k out | $2.00 | $10.00 | $2.00 | - |
| gpt-5.6-sol200k / 128k out | $3.50 | $18.00 | $0.35 | 60.9% |
| grok-4.6200k / 64k out | $1.70 | $5.00 | $0.40 | 60.9% |
| claude-sonnet-5200k / 64k out | $1.70 | $8.50 | $0.17 | 56.4% |
| gpt-5.6-terra200k / 128k out | $1.70 | $10.00 | $0.18 | 56.5% |
| gpt-5.6-luna200k / 128k out | $0.18 | $1.05 | $0.01 | 52.1% |
| gpt-5.5200k / 128k out | $4.50 | $27.00 | $0.45 | 55.8% |
| gemini-3.8-flash200k / 64k out | $1.00 | $4.50 | $0.25 | 58.7% |
| gemini-3.1-pro200k / 64k out | $1.75 | $10.50 | $0.45 | 51.2% |
| kimi-k3200k / 64k out | $2.60 | $13.00 | $0.55 | 59.7% |
| qwen3.8-max200k / 64k out | $1.70 | $5.20 | $0.40 | 58.1% |
| qwen3.8-2.4t-a95b200k / 64k out | $0.80 | $1.70 | $0.15 | 57.7% |
| deepseek-v4-pro200k / 64k out | $0.80 | $2.20 | $0.20 | 54.5% |
| deepseek-v4-flash200k / 32k out | $0.30 | $0.90 | $0.08 | 52.0% |
| glm-5.2200k / 128k out | $1.20 | $3.90 | $0.25 | 53.5% |
| qwen3.8-27b200k / 32k out | $0.22 | $0.80 | $0.05 | 52.0% |
| minimax-m3200k / 64k out | $0.27 | $1.05 | $0.08 | 47.5% |
| mimo-v2.5-pro200k / 64k out | $0.60 | $2.20 | $0.15 | 47.5% |
| agnes-2.5-flash200k / 64k out | $0.35 | $1.40 | $0.07 | 49.8% |
| deepseek-r1-0528-qwen3-8b96k / 32k out | $0.12 | $0.45 | $0.03 | 44.2% |
| gpt-oss-120b128k / 64k out | $0.45 | $1.80 | $0.09 | 50.6% |
Model lineup
GLM 5.3
Z.aiFlagship model by Z.ai. 200K context, always-on reasoning, top tier for code and agent tasks.
GLM-5.2
Z.aiPrevious generation of the GLM flagship. Still a strong coder for its price.
Agnes 2.5 Flash
Sapiens AISapiens AI's fast Flash-generation model. Low latency, 200K context, always-on reasoning.
Claude Fable 5.1
AnthropicIncremental update over Fable 5: better instruction following, long-horizon coding and tool use.
Claude Fable 5
AnthropicAnthropic's 2026 flagship. The strongest coder in the Claude family, 200K context, vision.
Claude Opus 5
AnthropicThe most capable Opus generation. Built for hard long-running tasks and deep research.
Claude Opus 4.8
AnthropicPrevious Opus generation, still a workhorse in production pipelines.
Claude Sonnet 5
AnthropicThe balanced workhorse of the Claude family: fast, strong coding, 200K context.
GPT-5.6 Sol
OpenAIFrontier tier of the GPT-5.6 family. Top-tier reasoning for hard coding and agents.
GPT-5.6 Terra
OpenAIBalanced mid tier of the GPT-5.6 family: frontier quality at a sane price.
GPT-5.6 Luna
OpenAIFast lightweight tier of the GPT-5.6 family. Ultra-cheap everyday workhorse.
GPT-5.5
OpenAIPrevious OpenAI flagship generation, still a reliable production workhorse.
GPT-OSS 120B
OpenAIOpenAI's open-weight GPT-OSS (120B MoE, ~5B active). Frontier-adjacent quality, open price.
Gemini 3.8 Flash
GoogleGoogle's fast Gemini: low latency, multimodal, 200K context, great price/quality.
Gemini 3.1 Pro
GoogleGoogle's Pro-tier Gemini: 200K context, complex reasoning and multimodal work.
Grok 4.6
xAIxAI's frontier model. Strong reasoning and coding with real-time knowledge.
Kimi K3
Moonshot AIMoonshot AI's flagship open-weight model. Top-tier coder and agent, 200K context.
Qwen3.8 Max
AlibabaAlibaba's strongest Qwen. Excellent coding and multilingual coverage, 200K context.
Qwen3.8 2.4T A95B
AlibabaOpen-weight Qwen MoE: 2.4T total, 95B active. Frontier quality at open prices.
Qwen3.8 27B
AlibabaCompact open Qwen (27B): fast, cheap and surprisingly capable.
DeepSeek V4 Pro
DeepSeekDeepSeek's flagship V4: reasoning-first coder with elite benchmark scores.
DeepSeek V4 Flash
DeepSeekDeepSeek's fast V4 variant: near-flagship quality, minimal latency and cost.
DeepSeek R1 0528 Qwen3 8B
DeepSeekDeepSeek's R1 (0528) reasoning distill on Qwen3-8B. Compact, fast, strong at math and logic.
MiniMax M3
MiniMaxMiniMax M3: fast long-context model tuned for agents and everyday coding.
MiMo V2.5 Pro
XiaomiXiaomi's MiMo Pro: budget-friendly reasoning model with solid coding chops.