Gemini 3 Flash Preview vs GPT-5 Mini
Gemini 3 Flash Preview and GPT-5 Mini both target the 'fast, cheap reasoning model' slot, but they land differently. Gemini 3 Flash is built for agentic workflows and multi-turn coding with near-Pro reasoning; GPT-5 Mini is a stripped-down GPT-5 aimed at lighter reasoning tasks with lower latency. The gap shows up in context window, output ceiling, and price — not in raw modality support, where Gemini adds audio and video input GPT-5 Mini lacks.
Spec vs spec
| Spec | Gemini 3 Flash Preview | GPT-5 Mini |
|---|---|---|
| Context window | 1.0M | 400K |
| Max output | 66K | 128K |
| Input modalities | text, image, file, audio, video | text, image, file |
| Output modalities | text | text |
| Knowledge cutoff | — | May 31, 2024 |
| Released | Dec 17, 2025 | Aug 7, 2025 |
| Reasoning | optional | always on |
Pricing
Per 1M tokens. Provider price plus the flat 3% fee — the sum is what you pay.
google/gemini-3-flash-preview
Input · 1M tokens
$0.500 + 3%$0.515
Output · 1M tokens
$3.00 + 3%$3.09
Cache read · 1M tokens
$0.050 + 3%$0.052
Cache write · 1M tokens
$0.083 + 3%$0.086
FEE — FLAT, EVERY MODEL3%
openai/gpt-5-mini
Input · 1M tokens
$0.250 + 3%$0.258
Output · 1M tokens
$2.00 + 3%$2.06
Cache read · 1M tokens
$0.025 + 3%$0.026
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-flash-preview
$11.33
$11.00 provider + 3%
openai/gpt-5-miniCheaper
$6.70
$6.50 provider + 3%
Benchmarks
Design Arena categories where both models have results. Higher Elo and lower rank win.
| Gemini 3 Flash Preview | GPT-5 Mini | |||
|---|---|---|---|---|
| Category | Elo | Rank | Elo | Rank |
| 3D | 1261 | #29 | 1114 | #73 |
| Code | 1239 | #33 | 1164 | #64 |
| Game dev | 1232 | #38 | 1194 | #54 |
| Websites | 1239 | #33 | 1167 | #62 |
Head-to-head preference voting. How we filter and rank
Pricing math
On OpenKey, Gemini 3 Flash Preview costs $0.515/M input and $3.09/M output tokens (provider price $0.50 / $3.00 plus the flat 3% fee). GPT-5 Mini runs $0.2575/M input and $2.06/M output (provider $0.25 / $2.00 + 3%). That's a 2.0x input price ratio in GPT-5 Mini's favor.
Run the numbers on a concrete workload — 10M input tokens + 2M output tokens — and Gemini 3 Flash Preview costs $11.00 total versus $6.50 for GPT-5 Mini. If you're processing high volumes of straightforward requests, that delta compounds fast. If you need the extra reasoning quality, it's worth paying for on a per-task basis rather than a blanket policy.
Coding and agentic performance
Gemini 3 Flash Preview has a dedicated 'agents' Design Arena bucket that GPT-5 Mini doesn't appear in, covering fullstack (rank 17, elo 1130), web apps (rank 17, elo 1185), mobile apps (rank 17, elo 1188), and Godot game dev (rank 8, elo 1218). In the shared 'models' code categories, Gemini 3 Flash Preview ranks 33rd (elo 1239, 57.6% win rate) versus GPT-5 Mini's 64th (elo 1164, 43.8% win rate). On game dev, Gemini ranks 38th (elo 1232) against GPT-5 Mini's 54th (elo 1194). Gemini leads on every comparable category here, which tracks with its design as an agentic/coding-focused model versus GPT-5 Mini's lighter-weight positioning.
Context and long documents
Gemini 3 Flash Preview supports 1,048,576 tokens of context against GPT-5 Mini's 400,000 — a 2.62x ratio. For anything involving large codebases, long transcripts, or multi-document retrieval without chunking, Gemini has meaningfully more room. GPT-5 Mini does have the higher max completion ceiling at 128,000 tokens versus Gemini's 65,535, so if your task is 'small input, very long generated output,' that flips in GPT-5 Mini's favor. Match the constraint to your workload shape rather than defaulting to whichever model has the bigger headline context number.
Modality differences
Gemini 3 Flash Preview accepts text, image, file, audio, and video input. GPT-5 Mini accepts text, image, and file only — no native audio or video input. Both output text only. If your pipeline needs to process audio transcriptions or video frames directly rather than pre-converting them, GPT-5 Mini is out and Gemini is the only option here.
When to pick each
Pick Gemini 3 Flash Preview when the task is agentic (multi-step tool use, coding agents), needs long context, or involves audio/video input. Pick GPT-5 Mini when you're running high-volume, lower-complexity reasoning at scale and cost per token matters more than top-end capability — its $6.50 cost on the 10M/2M workload versus Gemini's $11.00 is a real budget difference at scale. Both models are available on OpenKey through one API key, billed at provider list price plus a flat 3% fee.
Which model for which job
| Use case | Pick | Why |
|---|---|---|
| Coding agents / fullstack tasks | Gemini 3 Flash Preview | Ranks 17th on Design Arena fullstack (elo 1130) vs no comparable agent-bucket score for GPT-5 Mini |
| High-volume simple reasoning at scale | GPT-5 Mini | $6.50 vs $11.00 on a 10M-input/2M-output workload |
| Long-document or large-codebase analysis | Gemini 3 Flash Preview | 1,048,576 token context vs 400,000 — 2.62x more room |
| Audio or video input processing | Gemini 3 Flash Preview | Only Gemini supports audio and video input modalities |
| Very long generated output | GPT-5 Mini | 128,000 max completion tokens vs Gemini's 65,535 |
| Game dev / web app generation | Gemini 3 Flash Preview | Ranks 38th on gamedev (elo 1232) and 33rd on website (elo 1239) vs GPT-5 Mini's 54th and 62nd |
Questions
- Which is cheaper, Gemini 3 Flash Preview or GPT-5 Mini?
- GPT-5 Mini is cheaper on both input and output. On OpenKey it's $0.2575/M input and $2.06/M output, versus Gemini 3 Flash Preview's $0.515/M input and $3.09/M output — a 2.0x price ratio on input tokens in GPT-5 Mini's favor.
- How much would 10M input and 2M output tokens cost on each model?
- Gemini 3 Flash Preview costs $11.00 for that workload; GPT-5 Mini costs $6.50. That's roughly 41% less on GPT-5 Mini, worth factoring in if you're running this volume repeatedly.
- Which model has more context?
- Gemini 3 Flash Preview supports 1,048,576 tokens versus GPT-5 Mini's 400,000 — a 2.62x ratio. GPT-5 Mini does have a larger max output ceiling at 128,000 tokens versus Gemini's 65,535.
- Does either model support audio or video input?
- Only Gemini 3 Flash Preview does. Its input modalities include text, image, file, audio, and video, while GPT-5 Mini is limited to text, image, and file input.