MiniMax M2.7 High Speed
MiniMax M2.7 High Speed is the throughput-optimized variant of M2.7. It supports a context window of 204.8K tokens and a max output of 131.1K tokens.
import { streamText } from 'ai'
const result = streamText({ model: 'minimax/minimax-m2.7-highspeed', prompt: 'Why is the sky blue?'})Playground
Try out MiniMax M2.7 High Speed by MiniMax. Usage is billed to your team at API rates. Free users (those who haven't made a payment) get $5 of credits every 30 days.
MiniMax M2.7 High Speed
Ask MiniMax M2.7 High Speed anything to try it out.
Providers
Route requests across multiple providers. Copy a provider slug to set your preference. Visit the docs for more info. Using a provider means you agree to their terms, listed under Legal.
| Provider |
|---|
MiniMax | 205K | 0.9s | 48tps | $0.60/M | $2.40/M | Read:$0.06/M Write:$0.38/M | — | — | +1 | 03/18/2026 |
P50 throughput on live AI Gateway traffic, in tokens per second (TPS). Visit the docs for more info.
P50 time to first token (TTFT) on live AI Gateway traffic, in milliseconds. View the docs for more info.
Direct request success rate on AI Gateway and per-provider. Visit the docs for more info.
More models by MiniMax
| Model |
|---|
| 1M | 1.8s | 184tps | $0.60/M$0.30/M | $2.40/M$1.20/M | Read: $0.12/M$0.06/M Write: — | — | — | +2 |  | 05/31/2026 | |||
| 205K | 0.3s | 314tps | $0.15/M | $0.60/M | Read:$0.06/M Write:$0.38/M | — | — | +1 |  +2 | 03/18/2026 | |||
| 1M | 0.4s | 286tps | $0.07/M | $0.57/M | Read:$0.03/M Write:$0.38/M | — | — | +1 |   +3 | 02/12/2026 | |||
| 205K | 0.8s | 58tps | $0.60/M | $2.40/M | Read:$0.03/M Write:$0.38/M | — | — | +1 |  | 02/12/2026 | |||
| 205K | 0.5s | 241tps | $0.30/M | $1.20/M | Read:$0.03/M Write:$0.38/M | — | — | +1 | | 12/23/2025 | |||
| 205K | 0.7s | 92tps | $0.30/M | $1.20/M | Read:$0.03/M Write:$0.38/M | — | — | +1 | | 10/27/2025 |
About MiniMax M2.7 High Speed
Three architectural capabilities separate the 2.7 generation from earlier MiniMax releases. MiniMax M2.7 High Speed delivers all three at accelerated inference.
1. Agent-to-agent orchestration without middleware. Earlier MiniMax models operated as isolated workers. Coordinating them required external scaffolding: custom code to pass context, manage handoffs, and track dependencies. MiniMax M2.7 High Speed internalizes that orchestration layer. It manages context propagation, dependency resolution, and agent handoffs natively. In parallel architectures, compressing per-agent token generation shortens the critical path.
2. Runtime tool discovery. Prior generations consumed a static tool manifest declared at prompt time. MiniMax M2.7 High Speed breaks that constraint: it identifies, evaluates, and invokes tools dynamically as a task unfolds. For long-horizon automation where required actions can't be predicted upfront, this reduces the need to pre-enumerate every tool interaction.
3. Enterprise document processing. Structured data extraction, report synthesis, spreadsheet analysis, and document transformation join the capability set. A single endpoint now serves both engineering automation and business-process work, reducing the number of specialized models you manage.
Throughput remains high (see live metrics on this page). The generational leap is in what the model accomplishes per token, not how many tokens it produces.
What To Consider When Choosing a Provider
- Configuration: MiniMax M2.7 High Speed lists at roughly 2x the standard M2.7 input and output rates on many providers. AI Gateway's per-request cost tracking helps you quantify whether the throughput gain justifies the expense for your workload.
- Zero Data Retention: AI Gateway does not currently support Zero Data Retention for this model. See the documentation for models that support ZDR.
- Authentication: AI Gateway authenticates requests using an API key or OIDC token. You do not need to manage provider credentials directly.
When to Use MiniMax M2.7 High Speed
Best For
- Parallel agent architectures: Per-agent token velocity directly compresses end-to-end task completion
- Autonomous tool discovery: Workflows that must locate and invoke unfamiliar tools as subtasks emerge during execution
- Unified engineering and business: Pipelines that need code generation and document processing from one endpoint
- Native orchestration replacement: Organizations replacing external middleware with a model that coordinates agents natively
Consider Alternatives When
- Independent agents: Your agents never exchange context, so an earlier highspeed variant handles isolated coding at lower cost
- Batch jobs without pressure: Standard M2.7 produces identical results at the baseline rate
- Budget ceiling exceeded: The 2x per-token premium exceeds your budget regardless of latency benefit
Conclusion
MiniMax M2.7 High Speed adds agent orchestration, runtime tool discovery, and enterprise document work while sustaining the throughput that makes long-running, multi-agent sessions viable. It pairs the full M2.7 capability set with high-throughput inference for teams whose workloads have outgrown single-agent patterns.