GPT-5 Mini vs GPT-5.2
GPT-5 Mini and GPT-5.2 are both OpenAI models with identical 400K context windows and 128K max output, but they sit at opposite ends of the cost-vs-capability tradeoff. Mini launched August 2025 as the lightweight sibling of GPT-5; GPT-5.2 arrived December 2025 as OpenAI's frontier model with adaptive reasoning and agentic benchmarks Mini never ran. If you're picking between them for a real workload, the decision comes down to how much the elo gap is worth per token.
Spec vs spec
| Spec | GPT-5 Mini | GPT-5.2 |
|---|---|---|
| Context window | 400K | 400K |
| Max output | 128K | 128K |
| Input modalities | text, image, file | file, image, text |
| Output modalities | text | text |
| Knowledge cutoff | May 31, 2024 | — |
| Released | Aug 7, 2025 | Dec 10, 2025 |
| Reasoning | always on | optional |
Pricing
Per 1M tokens. Provider price plus the flat 3% fee — the sum is what you pay.
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%
openai/gpt-5.2
Input · 1M tokens
$1.75 + 3%$1.80
Output · 1M tokens
$14.00 + 3%$14.42
Cache read · 1M tokens
$0.175 + 3%$0.180
FEE — FLAT, EVERY MODEL3%
One workload, priced on both
10M input + 2M output tokens at each model's price, flat 3% fee included.
openai/gpt-5-miniCheaper
$6.70
$6.50 provider + 3%
openai/gpt-5.2
$46.87
$45.50 provider + 3%
Benchmarks
Design Arena categories where both models have results. Higher Elo and lower rank win.
| GPT-5 Mini | GPT-5.2 | |||
|---|---|---|---|---|
| Category | Elo | Rank | Elo | Rank |
| 3D | 1114 | #73 | 1156 | #60 |
| ASCII art | 1173 | #33 | 1199 | #22 |
| Code | 1164 | #64 | 1219 | #42 |
| Data viz | 1167 | #62 | 1245 | #31 |
| Game dev | 1194 | #54 | 1260 | #29 |
| SVG | 1149 | #42 | 1197 | #31 |
| UI components | 1161 | #57 | 1243 | #32 |
| Websites | 1167 | #62 | 1237 | #34 |
Head-to-head preference voting. How we filter and rank
Pricing math
On OpenKey, GPT-5 Mini costs $0.2575/M input and $2.06/M output (provider price $0.25 / $2.00 plus the flat 3% fee: 0.25 × 1.03 = 0.2575, 2.00 × 1.03 = 2.06). GPT-5.2 runs $1.8025/M input and $14.42/M output (1.75 × 1.03 = 1.8025, 14.00 × 1.03 = 14.42). On a 10M-input / 2M-output workload, Mini costs $6.50 total versus $45.50 for GPT-5.2 — a 7x difference. Input tokens alone are 7.14x cheaper on Mini (input price ratio 0.14, meaning GPT-5.2's input costs are roughly 1/0.14 = 7.14x Mini's). Both prices include OpenKey's flat 3% fee on top of provider list price — same key, same billing, for both models.
Coding and design benchmarks
On Design Arena's model-level categories, GPT-5.2 beats GPT-5 Mini in every category both models were tested on: codecategories (1219 vs 1164 elo, rank 42 vs 64), dataviz (1245 vs 1167, rank 31 vs 62), gamedev (1260 vs 1194, rank 29 vs 54), uicomponent (1243 vs 1161, rank 32 vs 57), asciiart (1199 vs 1173, rank 22 vs 33), svg (1197 vs 1149, rank 31 vs 42), 3d (1156 vs 1114, rank 60 vs 73), and website (1237 vs 1167, rank 34 vs 62). GPT-5.2 also has scores in five agent-arena categories — androidnative, fullstack, godotgamedev, mobileapps, webapps — where Mini has no benchmark data at all.
Reasoning and parameters
GPT-5 Mini makes reasoning mandatory, with efforts of high, medium, low, or minimal (default: medium) — you can't turn it off. GPT-5.2 makes reasoning optional, supporting xhigh, high, medium, low, or none (default: medium), which means you can dial it to zero for latency-sensitive calls or push to xhigh for harder problems. Both expose the same tool-use and structured-output parameters (tools, tool_choice, response_format, structured_outputs, seed), so switching between them in code is a model-ID change, not a rewrite.
Context and modality
Identical on paper: both support 400,000 tokens of context and cap output at 128,000 tokens (context ratio 1.0). Both accept text, image, and file input and return text only. The real difference is knowledge cutoff — GPT-5 Mini's is documented at 2024-05-31; GPT-5.2's isn't listed, consistent with it being the newer model (released 2025-12-10 vs Mini's 2025-08-07, a gap of about four months).
Which model for which job
| Use case | Pick | Why |
|---|---|---|
| High-volume chat or summarization | GPT-5 Mini | $6.50 vs $45.50 for the same 10M/2M workload |
| Agentic coding (fullstack, mobile, webapps) | GPT-5.2 | Only GPT-5.2 has agent-arena benchmarks; Mini has none |
| UI component or dataviz generation | GPT-5.2 | 1243 elo (uicomponent) and 1245 elo (dataviz) vs Mini's 1161 and 1167 |
| Latency-sensitive, simple tasks | GPT-5 Mini | Reasoning can be set to minimal; input cost is 7x lower |
| Tasks needing to disable reasoning entirely | GPT-5.2 | Supports effort 'none'; Mini's reasoning is mandatory |
| SVG or ASCII art generation | GPT-5.2 | 1197 elo (svg) and 1199 elo (asciiart) vs Mini's 1149 and 1173 |
Questions
- How much more does GPT-5.2 cost than GPT-5 Mini?
- On OpenKey, GPT-5.2 is $1.8025/M input and $14.42/M output versus Mini's $0.2575/M input and $2.06/M output. For a 10M-input/2M-output workload that's $45.50 versus $6.50 — GPT-5.2 costs about 7x more.
- Do both models support the same context length?
- Yes. Both GPT-5 Mini and GPT-5.2 support a 400,000-token context window and cap max completion output at 128,000 tokens — a context ratio of 1.0, so switching models doesn't change how much document you can feed in.
- Can I turn off reasoning on either model?
- Only on GPT-5.2, which supports a 'none' effort level alongside xhigh, high, medium, and low. GPT-5 Mini makes reasoning mandatory with a floor of 'minimal' — you can't disable it entirely.
- Which model wins on coding benchmarks?
- GPT-5.2 beats GPT-5 Mini in all eight shared Design Arena model categories, including codecategories (1219 vs 1164 elo) and gamedev (1260 vs 1194 elo). It also has five agent-arena scores Mini was never tested on.