MiniMax · 2026-10-19 · major
MiniMax M3.1 Flash — a fast agent model at $0.60 per million output tokens
MiniMax M3.1 Flash is MiniMax's new coding and agent model. MiniMax says it runs nearly three times faster online, at 200+ tokens per second, and cuts agent task cost by 70% versus its predecessor. It costs $0.15 / $0.60 per 1M tokens.
MiniMax's general release of M3.1 Flash: a cheap, fast model for coding, computer use and long agent tasks.
Key specs
| Deep swe v1.1 | 73.7% |
|---|---|
| Output speed | 200+ tokens/sec |
Quick facts
| Maker | MiniMax |
|---|---|
| Context window | 1M tokens |
| Availability | MiniMax API, M Plan, MiniMax Code |
| Weights | Promised later, not released yet |
| Speed (P50) | 170+ tokens/sec |
| Released | October 19, 2026 |
Benchmarks
| MiniMax M3.1 Flash | 73.7% | |
|---|---|---|
| DeepSeek V4.1 Flash 1017 | 77% | |
| Claude Opus 5 | 74% | |
| GPT-5.6 Sol | 73% | |
| Kimi K3 | 67.5% | |
| GLM-5.3 | 66.9% |
| MiniMax M3.1 Flash | 61.7% | |
|---|---|---|
| Claude Opus 5 | 70.6% | |
| GPT-5.6 Sol | 65.7% | |
| Kimi K3 | 58.3% | |
| DeepSeek V4.1 Flash | 54.3% |
Pricing
| Input | $0.15 / 1M tokens |
|---|---|
| Cached input | $0.015 / 1M tokens |
| Output | $0.60 / 1M tokens |
What is it?
MiniMax M3.1 Flash is the general release of the M3.1 Flash model that MiniMax first tested as a preview inside MiniMax Code. It is now on the MiniMax API as well as the M Plan subscription and MiniMax Code. MiniMax says model weights will be released later, and that a larger Pro model will follow.
How does it work?
To make inference faster, MiniMax says it used more aggressive sparsity. The company reports 200+ output tokens per second in everyday use and a median of 170+ tokens per second. The model card in MiniMax's docs lists a 1M-token context window and tunable thinking depth.
Why does it matter?
For teams that run many agent steps, cost per task matters more than a single top score. MiniMax's charts place M3.1 Flash near the low-cost end on Agents' Last Exam (29.6%) and Automation-Bench (62.8%), while its published table shows Claude Opus 5 still ahead on most coding and computer-use tests.
Who is it for?
developers running coding and computer-use agents on a budget
Frequently asked questions
- How much does MiniMax M3.1 Flash cost?
- MiniMax lists M3.1 Flash at $0.15 per million input tokens, $0.60 per million output tokens and $0.015 per million cached input tokens. MiniMax also says the cost of agentic tasks is 70% lower than with the model's predecessor, because the new version needs less spend per task.
- Are MiniMax M3.1 Flash weights open?
- Not yet. The MiniMax M3.1 Flash announcement says the model weights will be released later. For now the model is available through the MiniMax API, the M Plan subscription and the MiniMax Code coding agent, so it cannot be run locally on launch day.
- How does MiniMax M3.1 Flash compare with Claude Opus 5 and GPT-5.6 Sol?
- In MiniMax's own benchmark table, M3.1 Flash leads on Automation-Bench v1.0.6 (62.8) and LOCA-Bench 256K (60.3) and ties Claude Opus 5 on BankerToolBench (76.2). Claude Opus 5 scores higher on Terminal Bench 4.0, OSWorld 2.0, NL2Repo-Bench and HLE, and GPT-5.6 Sol leads on GDP.pdf AA.
- What changed from MiniMax M3.1-Flash-Preview?
- The M3.1-Flash-Preview was limited to the M Plan and MiniMax Code. MiniMax M3.1 Flash adds the pay-as-you-go MiniMax API with published per-token prices, and MiniMax says online inference is now nearly three times faster, at 200+ output tokens per second in everyday use.
Try it
MiniMax-M3.1-Flash on platform.minimax.io