Across 21 comparable benchmarks, GPT-5.2 has better reported results on 12 and worse on 9. 2 further shared results are listed below, each with the reason it is not counted.
Moderate evidence21 of 23 shared results countedHow we compare
Price basis: Standard rates. The price section below shows the applicable periods.
| Fact | GPT-5.2 | Gemini 3.5 Flash |
|---|---|---|
| Context window | 400K tokens | 1M tokens |
| Input / 1M | $1.75 | $1.50 |
| Output / 1M | $14.00 | $9.00 |
| Blended (4:1) | $3.07 | $2.03 |
| Status | Available | Deprecated |
| Released | December 11, 2025 | May 19, 2026 |
Blended is 4:1 input to output with 90% cache hits at listed rates — set your own mix.
Mean of 15 shared benchmarks grouped by capability on a 0–100 scale. How averages are built
GPT-5.2 ahead on 3 of 3 counted
Gemini 3.5 Flash ahead on 2 of 2 counted
GPT-5.2 ahead on 4 of 4 counted
Split 2–2 of 4 counted
Gemini 3.5 Flash ahead on 1 of 1 counted
GPT-5.2 ahead on 1 of 1 counted
Gemini 3.5 Flash ahead on 1 of 1 counted
Split 1–1 of 2 counted
No shared comparable results for Research & search, Cybersecurity.
Published scores, with matching modes where available. Gaps show GPT-5.2 relative to Gemini 3.5 Flash. Reading the results
| Benchmark | GPT-5.2 | Gemini 3.5 Flash | Gap |
|---|---|---|---|
| SWE-Bench Pro - PublicRepository Engineering · Accuracy | 55.6Extra-High · tools | 55.1≠Thinking High · tools | +0.5 pts#31 vs #34 of 66 |
| Terminal Bench HardAgentic Development · Accuracy | 47.0Extra-High · tools | 46.2≠Normal · tools | +0.8 pts#15 vs #18 of 157 |
| Vibe Code Bench v1.1Code Generation & Editing · Pass rate (%) | 53.5Extra-High · tools | 48.7≠Thinking High · tools | +4.8 pts#61 vs #65 of 121 |
| ARC-AGI-1Abstract Generalization · Score (%) | 90.5Deep Thinking · no tools · parallel | 92.5≠Thinking High · tools | −2.0 pts#32 vs #28 of 87 |
| ARC-AGI-2Abstract Generalization · Score (% solved); cost per task (USD)Not counted · Speed and cost telemetry is reported, not scored | 54.2Deep Thinking · no tools · parallel | 72.1≠Thinking High · tools | 0.75×#42 vs #27 of 80 |
| SimpleBenchCommonsense · Score (AVG@5) | 45.8Thinking High · no tools | 76.7≠Normal · no tools | 0.60×#71 vs #18 of 106 |
| Vending-Bench 2Logical Planning · Money Balance (USD)Not counted · Speed and cost telemetry is reported, not scored | 3591Unspecified Reasoning · tools | 5396Unspecified Reasoning · tools | 0.67×#47 vs #35 of 64 |
| AIME 2026Mathematics · Accuracy | 98.3Thinking High · no tools | 95.0Thinking High · no tools | +3.3 pts#5 vs #23 of 46 |
| FrontierMath Tier 4 v2Mathematics · Accuracy (verification_code) | 31.7Extra-High · tools | 26.8≠Thinking High · tools | +4.9 pts#36 vs #40 of 73 |
| FrontierMath v2Mathematics · Accuracy (verification_code) | 67.4Extra-High · tools | 62.8≠Thinking High · tools | +4.6 pts#34 vs #42 of 85 |
| Riemann-benchMathematics · Accuracy (%) | 37.6Extra-High · tools + internet | 36.8≠Thinking High · tools + web | +0.8 pts#17 vs #19 of 42 |
| CritPtScientific Reasoning · Score | 11.6Extra-High · no tools | 13.1≠Thinking High · no tools | 0.89×#50 vs #46 of 134 |
| GPQA DiamondScientific Reasoning · Accuracy | 93.2Deep Thinking · no tools · parallel | 92.8≠Thinking High · no tools | +0.4 pts#32 vs #36 of 200 |
| MedCodeClinical Workflows · Accuracy (%) | 49.8Extra-High · tools | 55.8≠Thinking High · tools | −6.1 pts#20 vs #6 of 97 |
| MedScribeClinical Workflows · Rubric pass rate (%) | 84.4Extra-High · tools | 76.6≠Thinking High · tools | +7.8 pts#35 vs #68 of 99 |
| MCP-AtlasTool Orchestration · Pass rate / claim coverage | 67.6Extra-High · tools | 83.6≠Thinking High · tools | −16.0 pts#42 vs #12 of 55 |
| EBR-benchMemory & Persistence · Score (% of 21 objectives, best of the two scored playthroughs) | 23.0Extra-High · tools | 4.8≠Thinking High · tools | +18.3 pts#19 vs #28 of 30 |
| MMMU-ProVisual Understanding · Accuracy | 80.4Thinking · no tools | 84.3≠Thinking High · no tools | −3.9 pts#29 vs #11 of 136 |
| SAGEService Workflows · Accuracy (%) | 49.3Extra-High · tools | 49.9≠Thinking High · tools | −0.6 pts#22 vs #18 of 82 |
| τ³-BankingService Workflows · Score | 32.2Thinking High · tools | 32.2Thinking High · tools | 1.00×#35 vs #36 of 120 |
| HLEKnowledge Exams · Accuracy | 45.5Deep Thinking · tools + web | 40.2≠Thinking High · tools | +5.3 pts#64 vs #77 of 151 |
| ECICapability Indices · ECI score (capability index, higher is better) | 153Reported best | 155Reported best | 0.99×#37 vs #32 of 168 |
| LiveBenchCross-capability Suites · Accuracy | 74.6Thinking High · no tools | 74.6Thinking High · no tools | −0.0 pts#36 vs #35 of 106 |
Repository Engineering · Accuracy
Agentic Development · Accuracy
Code Generation & Editing · Pass rate (%)
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)
Logical Planning · Money Balance (USD)
Not counted · Speed and cost telemetry is reported, not scored
Mathematics · Accuracy
Mathematics · Accuracy (verification_code)
Mathematics · Accuracy (verification_code)
Mathematics · Accuracy (%)
Scientific Reasoning · Score
Scientific Reasoning · Accuracy
Clinical Workflows · Accuracy (%)
Clinical Workflows · Rubric pass rate (%)
Tool Orchestration · Pass rate / claim coverage
Memory & Persistence · Score (% of 21 objectives, best of the two scored playthroughs)
Visual Understanding · Accuracy
Service Workflows · Accuracy (%)
Service Workflows · Score
A dash means no result is available.
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%
GPT-5.2 costs51%morethan Gemini 3.5 Flash per million tokens.Using standard rates at this token mix.
GPT-5.2: Standard pricing. OpenAI official pricing (developers.openai.com/api/docs/pricing) verified 2026-11-13; standard tier
Gemini 3.5 Flash: Standard pricing. Official Gemini Developer API standard input price. Official Gemini Developer API output price, including thinking tokens.
| Price / 1M tokens | GPT-5.2 | Gemini 3.5 Flash | Ratio |
|---|---|---|---|
| Input | $1.75 | $1.50 | 1.17× |
| Output | $14.00 | $9.00 | 1.56× |
| Cached input | $0.175 | $0.15 | 1.17× |
| Cache write | — | — | — |
| Blended | $3.07 | $2.03 | 1.51× |
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.
GPT-5.2 relative to Gemini 3.5 Flash
Summary: 1 of 3 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.
GPT-5.2
Gemini 3.5 Flash
Settings or sources differ or are not recorded; interpret the comparison with care.
GPT-5.2
Gemini 3.5 Flash
Settings or sources differ or are not recorded; interpret the comparison with care.
GPT-5.2
Gemini 3.5 Flash
| Benchmark | GPT-5.2 | Gemini 3.5 Flash |
|---|---|---|
| AutomationBench | Not reported | $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 |
| DeepSWE | Not reported | |
| Tax Agent Bench | Not reported | $0.8597 / task Average · Accuracy (%, rubric × citation quality): 21.6 Thinking High (With Tools) |
| Terminal-Bench 4.0 | Not reported | $5.98 / task Average · Resolution rate (%): 6.1 Unspecified Reasoning (With Tools) · vals-mini-swe |
| Vals Index v2.1 | Not reported | |
| WeirdML v3 | $9.2902 / task Average · Official score (0-1; higher is better): 0.05 Extra-High (With Tools) | Not reported |
| Attribute | GPT-5.2 | Gemini 3.5 Flash |
|---|---|---|
| Availability | Available | Deprecated |
| Provider | OpenAI | Google Deep Mind |
| Model type | Chat model | Multimodal model |
| First release | December 11, 2025 | May 19, 2026 |
| Context window | 400K tokens | 1M tokens |
| Max output | — | 64K tokens |
| Architecture | Undisclosed | Undisclosed |
| Documented capabilities | Reasoning, Multilingual | Reasoning, Multilingual, Function calling, Tool use, Structured output, Web search, Computer use, Long context |
| Knowledge cutoff | — | 2025-01 |
| Input modalities | Text, Image | Text, Image, Audio, Video |
| 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 GPT-5.2 or Gemini 3.5 Flash? 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.