Across 7 comparable benchmarks, GPT-5.2 Pro has better reported results on 5 and worse on 2. 1 further shared result is listed below, each with the reason it is not counted.
Moderate evidence7 of 8 shared results countedHow we compare
Price basis: Standard rates. The price section below shows the applicable periods.
| Fact | Gemini 3.5 Flash | GPT-5.2 Pro |
|---|---|---|
| Context window | 1M tokens | 256K tokens |
| Input / 1M | $1.50 | $21.00 |
| Output / 1M | $9.00 | $168.00 |
| Blended (4:1) | $2.03 | $50.40 |
| Status | Deprecated | Available |
| Released | May 19, 2026 | December 11, 2025 |
Blended is 4:1 input to output with 90% cache hits at listed rates — set your own mix.
Mean of 4 shared benchmarks grouped by capability on a 0–100 scale. How averages are built
Gemini 3.5 Flash ahead on 2 of 2 counted
GPT-5.2 Pro ahead on 2 of 2 counted
GPT-5.2 Pro ahead on 1 of 1 counted
No shared comparable results for Coding, Computer & tool use, Research & search, Long context, Multimodal, Professional work & productivity, Cybersecurity.
Published scores, with matching modes where available. Gaps show Gemini 3.5 Flash relative to GPT-5.2 Pro. Reading the results
| Benchmark | Gemini 3.5 Flash | GPT-5.2 Pro | Gap |
|---|---|---|---|
| ARC-AGI-1Abstract Generalization · Score (%) | 92.5Thinking High · tools | 90.5≠Thinking · no tools | +2.0 pts#28 vs #32 of 87 |
| ARC-AGI-2Abstract Generalization · Score (% solved); cost per task (USD)Not counted · Speed and cost telemetry is reported, not scored | 72.1Thinking High · tools | 54.2≠Thinking High · no tools | 1.33×#27 vs #43 of 80 |
| SimpleBenchCommonsense · Score (AVG@5) | 76.7Normal · no tools | 57.4≠Extra-High · no tools | 1.34×#18 vs #54 of 106 |
| FrontierMath Tier 4 v2Mathematics · Accuracy (verification_code) | 26.8Thinking High · tools | 46.0≠Extra-High · tools | −19.2 pts#40 vs #28 of 73 |
| FrontierMath v2Mathematics · Accuracy (verification_code) | 62.8Thinking High · tools | 74.0≠Extra-High · tools | −11.2 pts#42 vs #26 of 85 |
| GPQA DiamondScientific Reasoning · Accuracy | 92.8Thinking High · no tools | 93.2≠Thinking · no tools | −0.4 pts#36 vs #32 of 200 |
| HLEKnowledge Exams · Accuracy | 40.2Thinking High · tools | 50.0≠Thinking · tools | −9.8 pts#77 vs #50 of 151 |
| ECICapability Indices · ECI score (capability index, higher is better) | 155Reported best | 155Reported best | 0.99×#32 vs #25 of 168 |
Abstract Generalization · Score (%)
Abstract Generalization · Score (% solved); cost per task (USD)
Not counted · Speed and cost telemetry is reported, not scored
Commonsense · Score (AVG@5)
Mathematics · Accuracy (verification_code)
Mathematics · Accuracy (verification_code)
Scientific Reasoning · Accuracy
A dash means no result is available.
| Benchmark | Gemini 3.5 Flash | GPT-5.2 Pro |
|---|---|---|
| AIME 2026 | 95.0 | — |
| CritPt | 13.1 | — |
| FrontierMath - Tier 4 | — | 31.3 |
| ProofBench v1.1 (Lean 4) | 31.0 | — |
| Riemann-bench | 36.8 | — |
| SciCode | 53.9 | — |
| APEX-SWE | 36.1 | — |
| Code Migration | 26.8 | — |
| DeepSWE | 37.0 | — |
| MLE-Bench | 49.7 | — |
| SWE-Bench Pro - Public | 55.1 | — |
| Terminal Bench Hard | 46.2 | — |
| Terminal-Bench 4.0 | 6.6 | — |
| Vibe Code Bench v1.1 | 48.7 | — |
| APEX-Agents | 27.5 | — |
| MCP-Atlas | 83.6 | — |
| OSWorld-Verified | 78.4 | — |
| PinchBench v2 | 74.2 | — |
| Toolathlon-Verified | 67.3 | — |
| LiveBench | 74.6 | — |
| Vals Index | 53.1 | — |
| Vals Index v2.1 | 44.8 | — |
| Context Arena | 77.2 | — |
| EBR-bench | 4.8 | — |
| GDM-MRCR v2 (8-needle, 128K) | 77.3 | — |
| GDM-MRCR v2 (8-needle, 1M) | 26.6 | — |
| CharXiv RQ | 84.9 | — |
| GDP.pdf | 19.8 | — |
| GeoBench ACW | 85.0 | — |
| MMMU-Pro | 84.3 | — |
| AA-AnalystAgent | 45.0 | — |
| AutomationBench | 13.4 | — |
| EMB (Excel Modeling) | 63.5 | — |
| Finance Agent v2 | 57.9 | — |
| Harvey Lab-AA | 82.1 | — |
| Harvey's Legal Agent Benchmark | 2.5 | — |
| Legal Research Bench | 30.8 | — |
| Public Benefits Bench v1.1 | 59.5 | — |
| SAGE | 49.9 | — |
| Tax Agent Bench | 21.6 | — |
| τ³-Banking | 32.2 | — |
| Vending-Bench 2 | 5396 | — |
| BrowseComp | — | 77.9 |
| MedCode | 55.8 | — |
| MedScribe | 76.6 | — |
Time-of-day providers use peak-hour rates; fixed-price providers use standard rates. Off-peak discounts remain in the published details.
Drag the handle above the bar for cache hits, the one below it for the input : output mix · 5% steps
Blended price = effective input rate at 90% cache hits × 80% + output rate × 20%
Gemini 3.5 Flash costs96%lessthan GPT-5.2 Pro per million tokens — 24.9× cheaper.Using standard rates at this token mix.
Gemini 3.5 Flash: Standard pricing. Official Gemini Developer API standard input price. Official Gemini Developer API output price, including thinking tokens.
GPT-5.2 Pro: Standard pricing. OpenAI official pricing (developers.openai.com/api/docs/pricing) verified 2026-11-13; standard tier
| Price / 1M tokens | Gemini 3.5 Flash | GPT-5.2 Pro | Ratio |
|---|---|---|---|
| Input | $1.50 | $21.00 | 0.07× |
| Output | $9.00 | $168.00 | 0.05× |
| Cached input | $0.15 | — | — |
| Cache write | — | — | — |
| Blended | $2.03 | $50.40 | 0.04× |
Estimates use the selected token mix. Cache writes, tool fees and long-context surcharges are excluded.
Reported cost per task, paired with the score from the same run. Settings and sources may differ; costs do not affect the verdict.
Gemini 3.5 Flash relative to GPT-5.2 Pro
Summary: 1 of 2 shared benchmarks have matching recorded settings and sources. The average is a geometric mean of 1 positive-cost ratios, one equal weight per benchmark. This benchmark set can include related tests and indexes; it is not a workload cost estimate.
Effort, protocol, aggregation and source match; other run conditions may differ.
Gemini 3.5 Flash
GPT-5.2 Pro
Settings or sources differ or are not recorded; interpret the comparison with care.
Gemini 3.5 Flash
GPT-5.2 Pro
| Benchmark | Gemini 3.5 Flash | GPT-5.2 Pro |
|---|---|---|
| AIME 2026 | Not reported | |
| AutomationBench | $0.42 / task Average · Pass Rate: 13.4 Thinking High (With Tools) · zapier-private-heldout $0.38 / task Average · Pass Rate: 11.4 Thinking Medium (With Tools) · zapier-private-heldout 1 more settings | Not reported |
| DeepSWE | Not reported | |
| Tax Agent Bench | $0.8597 / task Average · Accuracy (%, rubric × citation quality): 21.6 Thinking High (With Tools) | Not reported |
| Terminal-Bench 4.0 | $5.98 / task Average · Resolution rate (%): 6.1 Unspecified Reasoning (With Tools) · vals-mini-swe | Not reported |
| Vals Index v2.1 | Not reported |
| Attribute | Gemini 3.5 Flash | GPT-5.2 Pro |
|---|---|---|
| Availability | Deprecated | Available |
| Provider | Google Deep Mind | OpenAI |
| Model type | Multimodal model | Reasoning model |
| First release | May 19, 2026 | December 11, 2025 |
| Context window | 1M tokens | 256K tokens |
| Max output | 64K tokens | — |
| Architecture | Undisclosed | Undisclosed |
| Documented capabilities | Reasoning, Multilingual, Function calling, Tool use, Structured output, Web search, Computer use, Long context | Reasoning |
| Knowledge cutoff | 2025-01 | — |
| Input modalities | Text, Image, Audio, Video | Text, Image |
| Output modalities | Text | Text |
| Weights | Closed / API only | Closed / API only |
| Commercial use | Proprietary | Proprietary |
These are reported-score comparisons, not controlled head-to-head tests.
What people found comparing these models, or using either one. Comments also appear on the models' own pages.
Used Gemini 3.5 Flash or GPT-5.2 Pro? Sign in with Google to share what worked, what did not, and which API or setup you tested.
No community comments on either model yet.
Alternatives with the most benchmark overlap for this matchup.