ModelsCompareBest forBenchmarksStatusPricingAPI
Compare / head-to-head

MiniMax M3vsQwen3.5 27B

MiniMax M3 leads 2 of 2 shared benchmarks. Qwen3.5 27B is 1.9x cheaper per token. MiniMax M3 has the larger context window (1M tokens).

Benchmarks from cited public sources; pricing from official pages; status from official provider feeds.

Shared benchmarks
2 – 0
MiniMax M3 leads
Cheaper per token
Qwen3.5 27B
1.9x cheaper, input + output
Larger context
MiniMax M3
1M tokens
Provider uptime (30d)
100% · —
MiniMax · Qwen
MiniMax M3
MiniMax · released 2026-06-01
textvision
Context
1M
Max out
Input /1M
$0.3
Output /1M
$1.2
Cached /1M
$0.06
Scores
3 · 3 core
textvision
Context
262K
Max out
Input /1M
$0.086
Output /1M
$0.688
Cached /1M
Scores
5 · 5 core
Quality

Benchmark matrix

BenchmarkMiniMax M3Qwen3.5 27BΔEdge
Reported by both · 2
GPQA Diamond90.9% epoch85.5% +5.4 ptMiniMax M3
LMArena Elo1433.5 1408.1 +25.4MiniMax M3
Only MiniMax M3 reports · 1
OTIS Mock AIME 2024-2025 (Epoch AI run)71.1% epochnot reported
Only Qwen3.5 27B reports · 3
SWE-bench Verifiednot reported75%
MMLU-Pronot reported86.1%
AIME 2026not reported91.7% matharena
Scores tagged "epoch" or "matharena" are independent runs, used only where the lab has not published its own; ⚠ marks rows MathArena flags as released after the competition. Higher is better on every row. Δ is MiniMax M3 minus Qwen3.5 27B in the benchmark's own unit. "Not reported" means the lab has not published that figure; it is not a zero. ↗ opens the source.
Specs & pricing

Side by side

SpecMiniMax M3Qwen3.5 27BEdge
Context window1M tokens262K tokensMiniMax M3
Max output
Input price / 1M$0.3 $0.086 Qwen3.5 27B
Output price / 1M$1.2 $0.688 Qwen3.5 27B
Cached input / 1M$0.06
Input + output / 1M
Lower is cheaper. List prices; batch, tool and regional fees excluded.
$1.5$0.774Qwen3.5 27B
Modalitiestext · visiontext · visionTie
Released2026-06-01
Cited benchmark scores35
Reliability

Provider status

All providers
More matchups

MiniMax M3 vs …

Built by Respan
Which one wins on your data?

Public benchmarks are a starting point. Run MiniMax M3 and Qwen3.5 27B on your own prompts with Respan evals, or route to either through one gateway key with automatic failover.