Gemini 3.5 Flash vs Gemini 3.1
Google's new Flash beats its own previous Pro tier on coding — at 25% lower cost. Here's exactly when to switch, and when 3.1 still earns its keep.
Gemini 3.5 Flash vs Gemini 3.1: side-by-side
| Gemini 3.5 Flash | Gemini 3.1 (Pro / Ultra) | |
|---|---|---|
| Released | 19 May 2026 | April 2026 |
| Context window | 1,048,576 tokens | Up to 2M tokens (Ultra) |
| Max output | 65,536 tokens | — |
| API price (per 1M) | $1.50 in / $9 out | ~$2 in / $12 out (Pro) |
| Cached input | $0.15 per 1M | — |
| Built for | Agentic coding, high-volume production | Multimodal scale — native video, audio, text |
| Standout | Beats 3.1 Pro on coding at lower cost | 2M context — largest available (Ultra) |
Benchmark comparison
| Benchmark | Gemini 3.5 Flash | Gemini 3.1 Pro | Winner |
|---|---|---|---|
| Terminal-Bench 2.1 | 76.2% | Below 3.5 Flash | 3.5 Flash |
| MCP Atlas | 83.6% | Below 3.5 Flash | 3.5 Flash |
| CharXiv Reasoning | 84.2% | Below 3.5 Flash | 3.5 Flash |
| GPQA | 82.8% | — | — |
Key takeaway
Independent trackers report 3.5 Flash outperforming 3.1 Pro on coding and agentic benchmarks while costing 25% less. When the newer, cheaper tier beats the older, pricier one, the migration decision mostly makes itself — context size is the one real exception.
Where Gemini 3.5 Flash wins
Coding and agents: the Terminal-Bench 2.1 and MCP Atlas results are the story — near-Pro (in fact, above-3.1-Pro) agentic capability at Flash economics. Cost: 25% cheaper on both input and output, and the $0.15/1M cached-input rate is a 10x discount that compounds in repeated-context agent loops.
Where Gemini 3.1 wins
Context: Gemini 3.1 Ultra's 2M-token window remains the largest available — double what 3.5 Flash offers. Multimodal scale: 3.1's headline was native video, audio and text in a single pass with no transcription step; for media-heavy pipelines it is still the specialized tool. (Google's incoming Gemini 3.5 Pro — 2M context — will eventually take over both roles.)
Which should you use?
- Agentic coding, tool orchestration → Gemini 3.5 Flash.
- High-volume production → Gemini 3.5 Flash — cheaper, better, heavy cache discounts.
- 2M-context document/media work → Gemini 3.1 Ultra, until 3.5 Pro ships broadly.
- Native video+audio pipelines → Gemini 3.1 Ultra.
Full Gemini 3.5 Flash overview Full Gemini 3.1 Ultra overview
Frequently asked questions
Is Gemini 3.5 Flash better than Gemini 3.1 Pro?
On coding and agentic benchmarks, yes — 76.2% Terminal-Bench 2.1, 83.6% MCP Atlas, 84.2% CharXiv Reasoning — while costing 25% less.
When should I stay on Gemini 3.1?
When you need the 2M-token context (Ultra) or heavyweight native video/audio multimodal processing. 3.5 Flash caps at ~1M context and 65,536 output tokens.
How much cheaper is 3.5 Flash?
25% on both sides vs 3.1 Pro: $1.50 vs $2.00 input, $9 vs $12 output, per million tokens. Cached input is $0.15/1M.
What about Gemini 3.5 Pro?
Announced as the new premium reasoning flagship with a 2M-token context, rolling out after Flash. Once pricing and benchmarks are public it becomes the natural top-tier comparison.