MiniMax M3 on QuickSilver Pro
MiniMax M3 is MiniMax's open-weight frontier model — a 1M-token context window, tuned for long-horizon agentic coding. It takes text input only. On QuickSilver Pro it's $0.24 input / $0.96 output per 1M tokens, matching OpenRouter's lowest current endpoint and 20% below its $0.30 / $1.20 catalog rate. It thinks by default; pass `reasoning: { enabled: false }` for direct non-thinking output.
At a glance
Open-weight frontier reasoning for long-horizon agentic coding over a 1M-token context.
Pricing comparison ($/1M tokens)
| Provider | Input | Output | vs QSP |
|---|---|---|---|
| QuickSilver Pro | $0.24 | $0.96 | lowest-cost |
| OpenRouter (minimax/minimax-m3) | $0.30 | $1.20 | 20% lower |
When to use
Reach for MiniMax M3 on long-horizon agentic coding: multi-step agents that plan, call tools, and iterate across a large repository within its 1M-token context. MiniMax positions it as an open-weight frontier model, so it suits tasks where reasoning quality matters but you still want to stay well under closed-model pricing.
When to use something else
For routine chat, short-context codegen, or single-shot tasks the reasoning overhead is unnecessary — DeepSeek V4 Flash ($0.112/$0.224) or MiMo-V2.5 ($0.112/$0.224) land those lower. For frontier multimodal reasoning at the top of the catalog, Kimi K3; for pure mathematical reasoning, DeepSeek V4 Pro.
Quickstart (curl)
curl https://api.quicksilverpro.io/v1/chat/completions \
-H "Authorization: Bearer $QSP_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "minimax-m3",
"messages": [{"role": "user", "content": "Hello!"}]
}'OpenAI-compatible. One-line migration via base_url.
FAQ
MiniMax positions M3 as an open-weight frontier model tuned for long-horizon agentic coding — planning, tool use, and iterating across a large repository within a 1M-token context. It accepts text input only, so pair it with a vision model if your agent must read screenshots. A/B it against DeepSeek V4 Pro and Kimi K3 on your own agent evals to see where it lands for your workload.
Yes — MiniMax M3 is an OpenAI-compatible chat completions endpoint on QuickSilver Pro. Set base_url=https://api.quicksilverpro.io/v1, paste your QSP key, and use model="minimax-m3". Streaming, tool calling, json_schema strict mode, and usage.cost accounting all work; M3 does not accept image input. M3 thinks by default; pass `reasoning: { enabled: false }` for non-thinking output.
QuickSilver Pro is $0.24 / $0.96 per 1M tokens. That is 20% below OpenRouter's $0.30 / $1.20 catalog rate and equal to its lowest current provider endpoint, so we do not claim to beat every possible OpenRouter route. Migration is a base_url + key swap, dropping the `minimax/` prefix.