Skip to main content

← Dracon Bench leaderboard

Model guide · Dracon Bench

minimax-m3.1-flash

Tested First seen 2026-11-18 (first probed by us) · First tested 2026-11-18 · NVIDIA NIM free endpoint (integrate.api.nvidia.com/v1)

MiniMax M3.1 Flash on NIM lands at 50.0 overall across five suites, rank 19. JSON is the bright spot: all five briefs conform. IFEval-50 50%, IFBench 20%, MMLU-Pro 2/10, GSM8K 60%; all five SVG briefs timed out, so this row has no SVG column. Several text prompts also needed timed-out retries before an answer landed; the final not_delivered count is 0. Pilot depth: 10 prompts per text suite and the 5 SVG briefs; every suite prints its own n and not_delivered.

Rank
#19
Overall
50.0
Suites
5 of 6
Ranked
#19 of 21

When the world heard about it

Timeline

  1. First reachable for us 2026-11-18 our first probe
  2. First tested by us 2026-11-18 Dracon Bench
  3. Officially announced 2026-10-19 llm-stats
  4. First official benchmark 2026-11-19 llm-stats: a public score existed by then — the first day we saw one
  • Reachable before we scored it same day
  • Announced before it reached the lane we served 30 days
  • A public score existed within 1 day

Track A · script-scored subsets

Scores

Subset scores for minimax-m3.1-flash.
SuiteScoreSamplePrecision
IFEval-50 prompt-strict50.0%5/10 prompts23.7%–76.3%
IFEval-50 instruction-strict52.9%9/17 constraints31.0%–73.8%
IFBench-50 prompt-strict20.0%2/10 prompts5.7%–51.0%
IFBench-50 instruction-strict18.2%2/11 constraints5.1%–47.7%
MMLU-Pro-5020.0%2/10 served5.7%–51.0%
JSON briefs conformance100.0%5/5 served56.6%–100.0%
GSM8K-5060.0%6/10 served31.3%–83.2%

The range beside a script-scored figure is a 95% interval over that figure's own sample: the prompts it followed, the constraints it met, or the items it answered. The instruction-strict figures are measured over CONSTRAINTS, which is why their sample line counts constraints rather than prompts. The judged total is not a proportion, so it has no interval: it carries the measured judge band instead: how far a re-judging of the same artifacts moved them, which is a different kind of number. The method page says what each one covers and what it does not.

Track A · per-type breakdown

What it failed

The instruction types these two suites check, and the checks the model missed. A type is checked once per prompt that carries it, so a type with seven checks was asked for seven times. The counts beside a type are that type's own; the suite's totals are in the scores table above. These are the failures behind the instruction-strict figure, which is the sum of them.

IFEval-50

Missed at least one check in 7 of the 15 instruction types checked.

The instruction types 7 of 15 the IFEval-50 suite checked that minimaxai/minimax-m3.1-flash missed, most missed first.
Instruction typeChecksMissed
startend · quotation22
change case · capital word frequency11
detectable format · json format11
detectable format · number highlighted sections11
length constraints · number paragraphs11
length constraints · number words11
punctuation · no comma21

IFBench-50

Missed at least one check in 7 of the 8 verifiers checked.

The verifiers 7 of 8 the IFBench-50 suite checked that minimaxai/minimax-m3.1-flash missed, most missed first.
VerifierChecksMissed
count · unique word count33
count · conjunctions11
count · person names11
count · word count range11
format · list11
words · keywords specific position11
words · words position11

Provider-reported usage

Thinking cost

What the run cost the model in thinking: provider-reported completion tokens per prompt, the thinking share, and the cost per correct answer. Reasoning-heavy models can exhaust fixed budgets before answering.

Thinking cost for minimax-m3.1-flash.
SuiteCostReading
IFEval-501,116/prompt76% thinking· 1,339/correct
GSM8K-50220/prompt64% thinking· 366/correct