MiniMax-M3

AA reports MiniMax-M3 around 55 and describes it as leading open weights once weights are released; Kilo lists 47.6% coding completion at $10.35 per attempt.

71 Floor score

confidence: medium

71 Floor Capability

Finished benchmark shape.

70 Operating Envelope

Moderate DSWE cost but slow average time.

77 Frontier Reality

AA Index 54.7 / Kilo #10 coding

67 Research Trust

Promising current/open-weight signal; provider and weight-release variance still need tracking.

Current signal

AA Index 54.7 / Kilo #10 coding

Evidence completeness

64%

5 score dimensions represented.

DSWE

20% pass@1

$5.57 / 57m / 98k output tokens

Calculation

Score components

70 Accuracy + grounding

Correct, grounded answers on ordinary verifiable tasks.

70 Instruction discipline

Follows exact user constraints and output formats.

66 Honesty + reliability

Avoids unsupported confidence and remains stable across runs.

74 Reasoning floor

Handles non-flashy reasoning without brittle failures.

70 Agent/coding

Completes tool, coding, and agent-style tasks when evidence exists.

82 Operating envelope

Cost, speed, token burn, verbosity, and friction.

Evidence

Sources used

Caveat

What to watch

Promising open-weight/current MiniMax row. Weight release and provider variance need tracking before treating it as a stable floor model.

Meaning

How to read the score

Each score is a 0-100 saturation scale. A score of 100 means that benchmark lane is complete for this model. Floor Capability is the finished benchmark shape; Operating Envelope, Frontier Reality, and Research Trust are provisional tabs that need prompt-pack backfill. Confidence is evidence quality, not another model grade. Evidence % is source coverage, not capability.

Citations

Benchmark sources

Artificial Analysis Intelligence Index Kilo Bench