Skip to main content

MiniMax M3

MiniMax-M3 is a multimodal foundation model from MiniMax. It supports text, image, and video inputs with text output, a 1M-token context window, and is suited for long-horizon agentic work, coding, and tool use. It is built on MiniMax Sparse Attention (MSA), which replaces full attention with KV-block selection to cut per-token compute at long context — roughly 1/20 the cost of the previous generation at 1M tokens, with substantially faster prefill and decode while retaining quality across most tasks. Trained as a native multimodal model on interleaved data and tuned for multi-turn, production-like collaboration via an interactive user-simulator framework, the model is oriented toward sustained, multi-step tasks rather than single-turn execution.

Model card on Hugging Face
Input / output priceUp to 50% off
$0.60 / $2.40$0.30 / $1.20 per 1M
Context
1M tokens
Released
Jun 1, 2026
Agent setup

Paste into Claude Code, Cursor, or Codex

Help me set up the Respan gateway: fetch and follow https://www.respan.ai/docs/agent-setup.md. Use model "minimax/minimax-m3".

Read docs

Providers

11 providers, cheapest first, OpenRouter last, priced per 1M tokens after gateway discounts.

ProviderInputOutput
EmpirioLabsempiriolabs/minimax-m3
$0.225/M$0.90/M
Concentrateconcentrate/minimax-m3
$0.28/M$1.10/M
DeepInfradeepinfra/minimax-m3
$0.28/M$1.10/M
Fireworksfireworks/minimax-m3
$0.30/M$1.20/M
MiniMax50% offminimax/minimax-m3
$0.60/M$0.30/M$2.40/M$1.20/M

Is a provider down? Fireworks AI status, MiniMax status, OpenRouter status

Pricing

Averaged across 11 providers, in USD per 1M tokens after gateway discounts.

Average input
$0.29 per 1M
Average output
$1.15 per 1M
Average cache read
$0.06 per 1M

Cost by volume

Total cost from 0 to 1T tokens on the cheapest and most expensive provider.

Highest$0.30/MFireworks, MiniMax, Nebius, Novita, Parasail, RunAnywhere, TogetherAI, and OpenRouterLowest$0.225/MEmpirioLabs

Throughput and latency

30-day averages from requests through the Respan gateway. Under each figure, every model on the gateway by its average, this one in blue.

Output speed

Higher is better

111tokens/sacross providers

Faster than 80% of the other 596 models on the gateway

Time to first token

Lower is better

1.3sacross providers

Faster than 58% of the other 153 models on the gateway

Response latency

Lower is better

18sacross providers

Faster than 11% of the other 776 models on the gateway

Benchmarks

The Respan Index plus 23 published scores from 9 sources, each linked to where it was published.

31.2/ 100

Respan Index#87 of 176

176 ranked models. The shaded band is this model's 95% interval, 27.2 to 36.1. Claude Sonnet 4.5 and Claude Opus 4.1 fall inside it, so the gap is within the margin of error.
MiniMax M3 by category: Coding 8.3, Agents 47.8, Reasoning 23.6, Knowledge & language 33.4, Human preference 53.0
  1. 87MiniMax M331.2
  2. 86Claude Sonnet 4.531.3
  3. 88Claude Opus 4.131.1

Published scores

BenchmarkScore
LiveBench CodingLiveBench68.2%
SWE-rebench 2026-05-15 to 2026-07-01Nebius47.2%
Terminal-Bench 4.0Vals AI1%

Compare MiniMax M3: MiniMax M3 vs MiniMax M2.7, MiniMax M3 vs MiniMax M3.1 Flash, MiniMax M3 vs MiniMax M2.1, MiniMax M3 vs MiniMax M2, MiniMax M3 vs MiniMax M1

Apps

Use MiniMax M3 in the tools you already run. Each guide points the app at the Respan gateway.

Setup guides for more apps are in the integration docs.

MiniMax M3 FAQ

Short answers, built from the data on this page.

MiniMax M3 is a model from MiniMax, released on Jun 1, 2026. MiniMax-M3 is a multimodal foundation model from MiniMax. It supports text, image, and video inputs with text output, a 1M-token context window, and is suited for long-horizon agentic work, coding, and tool use. It is built on MiniMax Sparse Attention (MSA), which replaces full attention with KV-block selection to cut per-token compute at long context — roughly 1/20 the cost of the previous generation at 1M tokens, with substantially faster prefill and decode while retaining quality across most tasks. Trained as a native multimodal model on interleaved data and tuned for multi-turn, production-like collaboration via an interactive user-simulator framework, the model is oriented toward sustained, multi-step tasks rather than single-turn execution.

On the Respan gateway, MiniMax M3 costs from $0.225 per 1M input tokens and $0.90 per 1M output tokens on EmpirioLabs, after gateway discounts. Averaged across its 11 providers, it costs $0.29 per 1M input tokens and $1.15 per 1M output tokens.

MiniMax M3 is available through 11 providers: EmpirioLabs, Concentrate, DeepInfra, Fireworks, MiniMax, Nebius, and 5 more.

MiniMax M3 accepts up to 1M tokens of context on its largest provider route.

Point any OpenAI-compatible SDK at https://gateway.respan.ai/v1 with your Respan API key and set the model to "minimax/minimax-m3". To set it up with a coding agent, copy the agent setup prompt at the top of this page.

MiniMax M3 scores 31.2 out of 100 on the Respan Index v1.1, ranking #87 of 176 models. By category: Coding 8.3, Agents 47.8, Reasoning 23.6, Knowledge & language 33.4, and Human preference 53.0. The Respan Index combines independent and lab-reported benchmark results into one score, using benchmark data collected by Respan.

MiniMax M3 has 23 published benchmark scores, including LiveBench Coding 68.2%, OSWorld-Verified 75.2%, and GPQA Diamond 90.9%. Each score links to where it was published; the data is collected by Respan.

Certified where it counts.

Built to meet the security and privacy standards that enterprise and healthcare teams require.

ISO 27001

The internationally recognized standard for information security management.

SOC 2

Secure, compliant management of your data across all of our systems.

GDPR

Operated under GDPR, the world's strictest standard for data privacy.

HIPAA

HIPAA compliant, with a BAA available for healthcare teams.

Works with your entire stack

Use Respan with your favorite frameworks and tools.

AnthropicAI21 LabsxAIVercel AI SDK
LangChainOpenRouterAssemblyAICursor
LiteLLMGroqBasetenMem0
CohereCogneeLlamaIndexFireworks
DeepSeekLinkupHaystackTogether AI
PerplexityInference.netPostHogRubyLLM
SlackMastraAzure OpenAIMistral
AWS BedrockMoonshotHyperspellBAML
ModaInstructorGoogle GeminiParasail
ReductoZapierAgnoNebius
OpenAINovita AIReplicateSuperagent

Built for AI agents.
Break less.
Ship more.