Gemini 3.1 Pro Preview vs MiniMax M3
Both models ship a 1,048,576-token context window and handle text plus image and video input, so the real decision comes down to price, output ceiling, and what each was tuned for. Gemini 3.1 Pro Preview (Google, released 2026-02-19) is a mandatory-reasoning frontier model. MiniMax M3 (MiniMax, released 2026-05-31) is optional-reasoning and built for long-horizon agent loops with a much larger 512,000-token max completion. The gap between them is mostly about how much you're willing to pay for a few extra intelligence points.
Spec vs spec
| Spec | Gemini 3.1 Pro Preview | MiniMax M3 |
|---|---|---|
| Context window | 1.0M | 1.0M |
| Max output | 66K | 512K |
| Input modalities | audio, file, image, text, video | text, image, video |
| Output modalities | text | text |
| Released | Feb 19, 2026 | May 31, 2026 |
| Reasoning | always on | optional |
Pricing
Per 1M tokens. Provider price plus the flat 3% fee — the sum is what you pay.
google/gemini-3.1-pro-preview
Input · 1M tokens
$2.00 + 3%$2.06
Output · 1M tokens
$12.00 + 3%$12.36
Cache read · 1M tokens
$0.200 + 3%$0.206
Cache write · 1M tokens
$0.375 + 3%$0.386
FEE — FLAT, EVERY MODEL3%
minimax/minimax-m3
Input · 1M tokens
$0.300 + 3%$0.309
Output · 1M tokens
$1.20 + 3%$1.24
Cache read · 1M tokens
$0.060 + 3%$0.062
FEE — FLAT, EVERY MODEL3%
One workload, priced on both
10M input + 2M output tokens at each model's price, flat 3% fee included.
google/gemini-3.1-pro-preview
$45.32
$44.00 provider + 3%
minimax/minimax-m3Cheaper
$5.56
$5.40 provider + 3%
Benchmarks
Design Arena categories where both models have results. Higher Elo and lower rank win.
| Gemini 3.1 Pro Preview | MiniMax M3 | |||
|---|---|---|---|---|
| Category | Elo | Rank | Elo | Rank |
| 3D | 1303 | #17 | 1306 | #16 |
| androidnative | 1059 | #25 | 990 | #27 |
| ASCII art | 1314 | #4 | 1219 | #15 |
| Code | 1290 | #18 | 1306 | #13 |
| Data viz | 1270 | #20 | 1295 | #11 |
| Game dev | 1264 | #26 | 1287 | #20 |
| SVG | 1347 | #2 | 1250 | #13 |
| UI components | 1322 | #8 | 1291 | #18 |
| Websites | 1294 | #15 | 1304 | #11 |
Head-to-head preference voting. How we filter and rank
Pricing math
Gemini 3.1 Pro Preview costs $2.00/M input and $12.00/M output from Google directly, or $2.06/M and $12.36/M through OpenKey (provider price × 1.03 for the flat 3% fee). MiniMax M3 runs $0.30/M input and $1.20/M output provider-side, $0.309/M and $1.236/M on OpenKey.
On a workload of 10M input tokens + 2M output tokens, Gemini costs **$44.00** on OpenKey while MiniMax M3 costs **$5.40** — an input price ratio of 6.67x. If you're running this workload at scale repeatedly, that gap compounds fast. Both models run through the same OpenKey key with the same flat 3% fee, so the price difference is entirely provider-side, not a platform markup.
Coding and agentic performance
On Artificial Analysis metrics, Gemini 3.1 Pro Preview scores 68.8 on coding index versus MiniMax M3's 58.6 — a real gap for pure code-generation tasks. Intelligence index is closer: 46.5 for Gemini vs 44.4 for MiniMax M3.
Agentic index flips the story: MiniMax M3 hits 35.4 vs Gemini's 21.4, meaning MiniMax M3 is tuned harder for multi-step tool-use loops relative to its other scores. In Design Arena's agentic categories, Gemini ranks well on agentic HTML slides (rank 5, 1226 elo) and Godot game dev (rank 6, 1220 elo), but MiniMax M3 only reports one agentic category (Android native, rank 27, 990 elo) — its agent-loop strength shows up in the AA agentic index, not in Design Arena's agent tasks specifically.
Context and output limits
Context length is identical at 1,048,576 tokens (context ratio 1.0), so neither model has an edge ingesting long documents. Max completion tokens is where they diverge hard: Gemini 3.1 Pro Preview caps output at 65,536 tokens, while MiniMax M3 allows up to 512,000 tokens of output.
For tasks that need to generate large volumes of text or code in one pass — full codebase rewrites, long transcripts, extensive structured output — MiniMax M3's higher ceiling matters more than raw intelligence score. For everything else, 65,536 tokens is plenty.
Modality and reasoning behavior
Gemini 3.1 Pro Preview accepts audio, file, image, text, and video input; MiniMax M3 accepts text, image, and video only — no audio or file input. If your pipeline needs audio transcription or file uploads, MiniMax M3 is out.
Reasoning is handled differently too. Gemini's reasoning is mandatory with three effort levels (high, medium, low; default medium) — you can't turn it off, which adds latency and token cost to every call. MiniMax M3's reasoning is optional, giving you control over when to pay for extended thinking versus a fast direct answer.
When to pick each
Pick Gemini 3.1 Pro Preview when coding quality is the bottleneck and you need the higher Artificial Analysis coding index (68.8) or top Design Arena SVG (rank 2) and ASCII art (rank 4) results. Pick MiniMax M3 when you're running agentic workflows at volume, need output beyond 65,536 tokens, or the 6.67x input price gap outweighs a few benchmark points. Both are available through OpenKey with the same key and the same flat 3% fee, so switching between them for A/B testing costs nothing but the token bill.
Which model for which job
| Use case | Pick | Why |
|---|---|---|
| High-volume agentic pipelines | MiniMax M3 | 6.67x cheaper on input tokens and higher agentic index (35.4 vs 21.4) |
| Complex coding tasks | Gemini 3.1 Pro Preview | Coding index 68.8 vs 58.6 |
| Long-form generation (>65K tokens out) | MiniMax M3 | 512,000 max completion tokens vs Gemini's 65,536 |
| Audio or file input processing | Gemini 3.1 Pro Preview | Only model of the two that accepts audio and file inputs |
| Budget-constrained batch jobs | MiniMax M3 | $5.40 vs $44.00 for the same 10M-in/2M-out workload |
| SVG or ASCII-art generation | Gemini 3.1 Pro Preview | Design Arena rank 2 on SVG and rank 4 on ASCII art |
Questions
- How much does each model cost on OpenKey?
- Gemini 3.1 Pro Preview is $2.06/M input and $12.36/M output on OpenKey (provider price $2.00/$12.00 plus the flat 3% fee). MiniMax M3 is $0.309/M input and $1.236/M output (provider price $0.30/$1.20 plus 3%). A 10M-in/2M-out workload costs $44.00 on Gemini versus $5.40 on MiniMax M3.
- Which model has a bigger context window?
- Neither — both support 1,048,576 tokens of context, a 1.0 context ratio. The difference is on the output side: MiniMax M3 can generate up to 512,000 tokens per response while Gemini 3.1 Pro Preview caps out at 65,536.
- Which model scores higher on coding benchmarks?
- Gemini 3.1 Pro Preview, with an Artificial Analysis coding index of 68.8 versus MiniMax M3's 58.6. Gemini also ranks higher in several Design Arena code-related categories, including rank 2 on SVG generation.
- Does either model require reasoning to be enabled?
- Yes — Gemini 3.1 Pro Preview has mandatory reasoning with three effort levels (high, medium, low, default medium), so you can't disable it. MiniMax M3's reasoning is optional, giving you more control over latency and cost per call.