Quick Take

MiniMax-M3 and Google’s Gemini 3.5 Flash (medium) represent two distinct approaches to model deployment. Released on June 1, 2026, MiniMax-M3 enters the market with a zero-cost pricing structure. Gemini 3.5 Flash (medium), released slightly earlier on May 19, 2026, operates on a paid tier but provides detailed performance data regarding token speed and latency.

Benchmark Read

Both models demonstrate high levels of capability across various metrics. MiniMax-M3 holds an Intelligence index of 54.7 and a Coding index of 43.4. Gemini 3.5 Flash (medium) leads slightly with an Intelligence index of 54.8 and a Coding index of 43.9.

In specific benchmarks, MiniMax-M3 outperforms Gemini 3.5 Flash in:

• GPQA: 0.929 vs 0.921
• IFBench: 0.8286 vs 0.7456
• LCR: 0.74 vs 0.71
• TerminalBench Hard: 0.4242 vs 0.3939

Conversely, Gemini 3.5 Flash (medium) shows stronger results in:

• HLE: 0.399 vs 0.371
• SciCode: 0.53 vs 0.454
• TAU2: 0.9561 vs 0.8889

Math index data is currently unknown for both models.

Cost and Speed

Cost is the primary differentiator. MiniMax-M3 is priced at $0.00 per 1M tokens for both input and output. In contrast, Gemini 3.5 Flash (medium) costs $1.50 per 1M input tokens and $9.00 per 1M output tokens, resulting in a blended cost of $3.38 per 1M tokens.

Regarding performance, Gemini 3.5 Flash (medium) provides clear metrics: an output speed of 206.494 tokens per second and a time to first token of 11.76 seconds. Corresponding performance data for MiniMax-M3 remains unknown.

Best Fit

MiniMax-M3 is best suited for users or developers looking to minimize infrastructure costs while maintaining high-level performance in logic and instruction following. Gemini 3.5 Flash (medium) is better suited for enterprise applications where performance guarantees, speed metrics, and integration with Google’s development tools are required.