Is MiMo v2.6 Pro available with Zero Data Retention on Elosia?

MiMo v2.6 Pro runs on Elosia's standard backend; ZDR is available on many other models in the catalog.

How much does MiMo v2.6 Pro cost on Elosia?

MiMo v2.6 Pro is billed by usage at $0.435 per 1M input tokens and $0.87 per 1M output tokens, within your Elosia plan.

What is MiMo v2.6 Pro's context window?

MiMo v2.6 Pro supports a context window of up to 1,048,576 tokens.

Is MiMo v2.6 Pro hosted in the EU?

MiMo v2.6 Pro is not available on Elosia's EU-hosted (EdenAI) backend; it runs on the default backend.

When should you choose MiMo v2.6 Pro?

MiMo v2.6 Pro is a flagship model, best for the most demanding reasoning, agentic and long-context tasks.

MiMo v2.6 Pro

XiaomiFlagship

ThinkingTool UseVisionStructured Output

About this model

MiMo-V2.6-Pro is the flagship foundation model developed by Xiaomi. Built at a scale of over 1T parameters, it is designed to push the ceiling of capability for the most demanding...

Performance Tier

Flagship

MiMo v2.6 Pro is a flagship model from Xiaomi : the most capable in their lineup.

Best-in-class model from this provider. Highest performance across benchmarks, ideal for demanding tasks.

Pricing

This model is included in Elosia plans

Affordable

Low cost. Suitable for sustained use and high-volume interactions.

Type	per 1M tokens
Input (prompt)	$0.435
Output (completion)	$0.870
Cache read	$0.0036

Capabilities

Context Length1.0M

Max Output Tokens131K

TokenizerOther

Inputtext, image, video, audio

Outputtext

Release DateSeptember 21, 2026

Benchmarks

General Intelligence

MMLU

Not reported

GPQA Diamond

Not reported

Mathematics

MATH-500

Not reported

Programming

HumanEval

Not reported

SWE-bench Verified

Not reported

Reasoning

IFEval

Not reported

Humanity's Last Exam

49.4%

Agentic

Terminal-Bench 2.1

89.9%

Where does MiMo v2.6 Pro stand?

Compare its performance index against every other model.

View the performance leaderboard

Recommended Use Cases

CodingAnalysisResearchData Extraction

Strengths

Sparse MoE with 1.02T total and 42B active parameters, open weights under the permissive MIT license
Native omni-modal input: text, image, video and audio in a single model over a 1M-token context
Trained with one mixed reinforcement-learning run across coding, general agents, visual tasks and cybersecurity, so strategies transfer to agent harnesses not seen in training
Group-wise agentic grading during training steers the model toward shorter solution paths and fewer tokens per task
Built-in multi-token prediction drafter that proposes several tokens per forward pass for speculative decoding

Limitations

The vendor publishes almost no standard academic benchmarks (no GPQA Diamond, SWE-bench or MMMU-Pro); the Humanity's Last Exam score shown comes from third-party testing
Tends to answer rather than abstain on questions it does not know, according to independent factual-reliability testing
Slow generation on the standard endpoint; low-latency use relies on a separate speed-optimized edition
Self-hosting the full model requires a multi-node GPU cluster