modelbench日本語

Gemini 3 Flash Preview

GooglePreview

Input price
$0.50
Per 1 million tokens (USD)
Output price
$3.00
Per 1 million tokens (USD)
Effort levels
4total
Default: High
Benchmark results
20total
Benchmark types: 18

Model information

API model ID
gemini-3-flash-preview
release date
Unknown
effort
Minimal / Low / Med / High (default)
Effort source
Based on the table in Gemini API thinking documentation (aistudio.google.com/docs/thinking, retrieved 2026-11-13): gemini-3-flash-preview is On (high); levels are minimal, low, medium, high.
Claude Code
unavailable
Codex
unavailable
Availability notes
Support in Claude Code / Codex is not explicitly stated officially (Gemini API, AI Studio, Antigravity, etc. are official routes). The official model list for Claude Code / Codex (checked 2026-11-12) contains only Anthropic / OpenAI models, respectively, so both are false.
Availability source
https://aistudio.google.com/docs/models

Price (USD per 1 million tokens)

Prices do not vary by effort; token consumption does.

input$0.50
output$3.00
cache read$0.05
cache write—

audio input $1.00 / audio cache $0.10, storage $1.00/hr. Efforts not collected because they are not stated on the thinking page

Gemini 3 Flash Preview gemini-3-flash-preview $0.50 / $3.00 / $0.05

Source: https://aistudio.google.com/docs/pricing

Benchmark results

BenchmarkeffortScoreSource cost (metric)measured/reported bysourceConditions and extraction method
Terminal-Bench 2.1not statednot stated58.0%—GoogleGemini 3.5 Flash model cardofficial · Retrieved 2026-11-13Vendor-reported value. Table in the Google model card. methodology PDF (deepmind.google/models/evals-methodology/gemini-3-5-flash): All of the results are all run with the Gemini API ... with default sampling settings unless indicated otherwise. 3.5 Flash and 3 Flash computed in-house with the Terminus 2 harness (3.1 Pro is not used because its value is from the leaderboard). Column heading is Gemini 3 Flash (gemini-3-flash-preview is the only match in the Gemini API model list). thinking level not statedtranscribed from table
Source excerptCoding | Terminal-bench 2.1 Agentic terminal coding | Terminus-2 harness | 58.0%Location Gemini 3.5 Flash model card Evaluation table, Terminal-bench 2.1 row, Gemini 3 Flash column. In-house computation basis: methodology PDF, Coding / Terminal-Bench 2.1
CharXiv Reasoningnot statednot stated80.3%—GoogleGemini 3.5 Flash model cardofficial · Retrieved 2026-11-13Vendor-reported value. Table in the Google model card. methodology PDF (deepmind.google/models/evals-methodology/gemini-3-5-flash): All of the results are all run with the Gemini API ... with default sampling settings unless indicated otherwise. Gemini computed in-house; without tools. Column heading is Gemini 3 Flash (gemini-3-flash-preview is the only match in the Gemini API model list). thinking level not statedtranscribed from table
Source excerptMultimodal | CharXiv Reasoning Information synthesis from complex charts | No tools | 80.3%Location Gemini 3.5 Flash model card Evaluation table, CharXiv Reasoning(No tools) row, Gemini 3 Flash column. In-house computation basis: methodology PDF, Image / CharXiv Reasoning
GPQA Diamondnot statednot stated83.2%—Epoch AIEpoch AI, Capabilities & benchmarking (benchmark_data.zip, gpqa_diamond.csv) CC BY 4.0third-party · Retrieved 2026-11-12Run by Epoch AI (third-party measurement). Source label "Gemini 3 Flash Preview". No effort suffix in Model version. Data retrieved 2026-11-12. No effort because effort is not stated in the source.transcribed from table
Source excerptgpqa_diamond.csv: Model version=gemini-3-flash-preview, Best score (across scorers)=0.8320707070707071, stderr=0.019036353018587762, Started at=2025-12-17T22:39:00.278ZLocation gpqa_diamond.csv, data row 195 (file line 277), id=oAu9CZSNPZgwiwKnoNpM8Q, Model version=gemini-3-flash-preview
GPQA DiamondHigh (default)High89.4%—Epoch AIEpoch AI, Capabilities & benchmarking (benchmark_data.zip, gpqa_diamond.csv) CC BY 4.0third-party · Retrieved 2026-11-12Run by Epoch AI (third-party measurement). Source label "Gemini 3 Flash Preview". effort is indicated by the suffix _high in Epoch's Model version. Data retrieved 2026-11-12.transcribed from table
Source excerptgpqa_diamond.csv: Model version=gemini-3-flash-preview_high, Best score (across scorers)=0.8939393939393939, stderr=0.021938047738853106, Started at=2026-08-06T22:22:58.000ZLocation gpqa_diamond.csv, data row 132 (file line 192), id=gVdraDT3pXmaExP9VbqME4, Model version=gemini-3-flash-preview_high
FrontierMath Tiers 1-3(v2)not statednot stated51.2%—Epoch AIEpoch AI, Capabilities & benchmarking (benchmark_data.zip, frontiermath_tiers_1_3_v2.csv) CC BY 4.0third-party · Retrieved 2026-11-12Run by Epoch AI (third-party measurement). Source label "Gemini 3 Flash Preview". No effort suffix in Model version. Data retrieved 2026-11-12. No effort because effort is not stated in the source.transcribed from table
Source excerptfrontiermath_tiers_1_3_v2.csv: Model version=gemini-3-flash-preview, Best score (across scorers)=0.512280701754386, stderr=0.02966059084324674, Started at=2026-06-11T23:44:30.000ZLocation frontiermath_tiers_1_3_v2.csv, data row 109 (file line 139), id=EtgrDq6ei6kZe8K7QbXUe3, Model version=gemini-3-flash-preview
FrontierMath Tier 4(v2)not statednot stated17.1%—Epoch AIEpoch AI, Capabilities & benchmarking (benchmark_data.zip, frontiermath_tier_4_v2.csv) CC BY 4.0third-party · Retrieved 2026-11-12Run by Epoch AI (third-party measurement). Source label "Gemini 3 Flash Preview". No effort suffix in Model version. Data retrieved 2026-11-12. No effort because effort is not stated in the source.transcribed from table
Source excerptfrontiermath_tier_4_v2.csv: Model version=gemini-3-flash-preview, Best score (across scorers)=0.17073170731707318, stderr=0.05949419959829497, Started at=2026-06-11T23:44:30.000ZLocation frontiermath_tier_4_v2.csv, data row 65 (file line 102), id=ZczEhu69FQ9Cnns8NT6ew4, Model version=gemini-3-flash-preview
OTIS Mock AIME 2024-2025not statednot stated92.8%—Epoch AIEpoch AI, Capabilities & benchmarking (benchmark_data.zip, otis_mock_aime_2024_2025.csv) CC BY 4.0third-party · Retrieved 2026-11-12Run by Epoch AI (third-party measurement). Source label "Gemini 3 Flash Preview". No effort suffix in Model version. Data retrieved 2026-11-12. No effort because effort is not stated in the source.transcribed from table
Source excerptotis_mock_aime_2024_2025.csv: Model version=gemini-3-flash-preview, Best score (across scorers)=0.9277777777777778, stderr=0.035196259111492545, Started at=2025-12-17T22:40:30.896ZLocation otis_mock_aime_2024_2025.csv, data row 202 (file line 289), id=H3E7EjjXfGCPnm9ooLdmAo, Model version=gemini-3-flash-preview
OTIS Mock AIME 2024-2025High (default)High95.6%—Epoch AIEpoch AI, Capabilities & benchmarking (benchmark_data.zip, otis_mock_aime_2024_2025.csv) CC BY 4.0third-party · Retrieved 2026-11-12Run by Epoch AI (third-party measurement). Source label "Gemini 3 Flash Preview". effort is indicated by the suffix _high in Epoch's Model version. Data retrieved 2026-11-12.transcribed from table
Source excerptotis_mock_aime_2024_2025.csv: Model version=gemini-3-flash-preview_high, Best score (across scorers)=0.9555555555555556, stderr=0.031067790907534743, Started at=2026-08-06T22:25:57.000ZLocation otis_mock_aime_2024_2025.csv, data row 140 (file line 205), id=a4uE6KgMhGSUjcnemRnbkA, Model version=gemini-3-flash-preview_high
SimpleQA VerifiedHigh (default)High66.8%—Epoch AIEpoch AI, Capabilities & benchmarking (benchmark_data.zip, simpleqa_verified.csv) CC BY 4.0third-party · Retrieved 2026-11-12Run by Epoch AI (third-party measurement). Source label "Gemini 3 Flash Preview". effort is indicated by the suffix _high in Epoch's Model version. Data retrieved 2026-11-12.transcribed from table
Source excerptsimpleqa_verified.csv: Model version=gemini-3-flash-preview_high, Best score (across scorers)=0.668, stderr=0.014899597242811567, Started at=2026-08-27T19:30:06.000ZLocation simpleqa_verified.csv, data row 63 (file line 110), id=7cc6cDCXjsPjtKt36bbeMR, Model version=gemini-3-flash-preview_high
Chess Puzzles(Epoch AI)High (default)High40.0%—Epoch AIEpoch AI, Capabilities & benchmarking (benchmark_data.zip, chess_puzzles.csv) CC BY 4.0third-party · Retrieved 2026-11-12Run by Epoch AI (third-party measurement). Source label "Gemini 3 Flash Preview". effort is indicated by the suffix _high in Epoch's Model version. Data retrieved 2026-11-12.transcribed from table
Source excerptchess_puzzles.csv: Model version=gemini-3-flash-preview_high, Best score (across scorers)=0.4, stderr=0.0492365963917331, Started at=2026-10-13T13:42:10.000ZLocation chess_puzzles.csv, data row 104 (file line 158), id=UUNZJtgib9XbCjLEdhtGaT, Model version=gemini-3-flash-preview_high
Epoch Capabilities Index(ECI)not statednot stated152.32—Epoch AIEpoch AI, epoch_capabilities_index/eci_scores.csv (CC BY 4.0)third-party · Retrieved 2026-11-12A per-model metric that combines results across multiple benchmarks and multiple efforts (not per effort). Calculated by Epoch AI.transcribed from table
Source excerpteci_scores.csv: Model=Gemini 3 Flash, eci=152.32, eci_ci_low=150.75, eci_ci_high=154.03Location epoch_capabilities_index/eci_scores.csv, data row 59 (file line 60), Model=Gemini 3 Flash
SWE-Bench Pronot statednot stated49.6%—GoogleGemini 3.5 Flash model cardofficial · Retrieved 2026-11-13Vendor-reported value. Table in the Google model card. methodology PDF (deepmind.google/models/evals-methodology/gemini-3-5-flash): All of the results are all run with the Gemini API ... with default sampling settings unless indicated otherwise. Gemini computed in-house; average of 5 runs; internal version of the Antigravity harness; single trial. Column heading is Gemini 3 Flash (gemini-3-flash-preview is the only match in the Gemini API model list). thinking level not statedtranscribed from table
Source excerptSWE-Bench Pro (Public) Diverse agentic coding tasks | Single attempt | 49.6%Location Gemini 3.5 Flash model card Evaluation table, SWE-Bench Pro (Public) row, Gemini 3 Flash column. In-house computation basis: methodology PDF, Coding / SWE-Bench Pro
OSWorld-Verifiednot statednot stated65.1%—GoogleGemini 3.5 Flash model cardofficial · Retrieved 2026-11-13Vendor-reported value. Table in the Google model card. methodology PDF (deepmind.google/models/evals-methodology/gemini-3-5-flash): All of the results are all run with the Gemini API ... with default sampling settings unless indicated otherwise. Gemini computed in-house; average of 5 runs (1 trial each); up to 100 steps. Column heading is Gemini 3 Flash (gemini-3-flash-preview is the only match in the Gemini API model list). thinking level not statedtranscribed from table
Source excerptUI Control | OSWorld-Verified Agentic computer use | 65.1%Location Gemini 3.5 Flash model card Evaluation table, OSWorld-Verified row, Gemini 3 Flash column. In-house computation basis: methodology PDF, UI Control / OSWorld-Verified
MMMU-Pronot statednot stated81.2%—GoogleGemini 3.5 Flash model cardofficial · Retrieved 2026-11-13Vendor-reported value. Table in the Google model card. methodology PDF (deepmind.google/models/evals-methodology/gemini-3-5-flash): All of the results are all run with the Gemini API ... with default sampling settings unless indicated otherwise. Gemini computed in-house; average of Standard (10 choices) and Vision; without tools. Column heading is Gemini 3 Flash (gemini-3-flash-preview is the only match in the Gemini API model list). thinking level not statedtranscribed from table
Source excerptMMMU-Pro Multimodal understanding and reasoning | No tools | 81.2%Location Gemini 3.5 Flash model card Evaluation table, MMMU-Pro(No tools) row, Gemini 3 Flash column. In-house computation basis: methodology PDF, Image / MMMU-Pro
MRCR v2 (8-needle, 128k)not statednot stated67.2%—GoogleGemini 3.5 Flash model cardofficial · Retrieved 2026-11-13Vendor-reported value. Table in the Google model card. methodology PDF (deepmind.google/models/evals-methodology/gemini-3-5-flash): All of the results are all run with the Gemini API ... with default sampling settings unless indicated otherwise. All models computed in-house; cumulative score up to 128k. Column heading is Gemini 3 Flash (gemini-3-flash-preview is the only match in the Gemini API model list). thinking level not statedtranscribed from table
Source excerptLong context | MRCR v2 (8-needle) Long context performance | 128k (average) | 67.2%Location Gemini 3.5 Flash model card Evaluation table, MRCR v2 (8-needle) 128k (average) row, Gemini 3 Flash column. In-house computation basis: methodology PDF, Long Context / MRCR v2
MRCR v2 (8-needle, 1M)not statednot stated22.1%—GoogleGemini 3.5 Flash model cardofficial · Retrieved 2026-11-13Vendor-reported value. Table in the Google model card. methodology PDF (deepmind.google/models/evals-methodology/gemini-3-5-flash): All of the results are all run with the Gemini API ... with default sampling settings unless indicated otherwise. All models computed in-house; pointwise value at 1M. Column heading is Gemini 3 Flash (gemini-3-flash-preview is the only match in the Gemini API model list). thinking level not statedtranscribed from table
Source excerpt1M (pointwise) | 22.1%Location Gemini 3.5 Flash model card Evaluation table, MRCR v2 (8-needle) 1M (pointwise) row, Gemini 3 Flash column. In-house computation basis: methodology PDF, Long Context / MRCR v2
τ²-bench airline (v1.0.1)High (default)High82.5%—Sierraτ²-bench leaderboard submission gemini-3-flash_sierra_2026-03-02 (sierra-research/tau2-bench, MIT)third-party · Retrieved 2026-11-13τ²-bench leaderboard submission gemini-3-flash_sierra_2026-03-02 (submission_type standard, tau2_bench_version 1.0.1). Submitter "Sierra", model_organization "Google": Run by Sierra, which operates τ²-bench (third-party measurement). Verified on the leaderboard (trajectories available, prompts not modified, no questions omitted). The value is pass^1 (%). reasoning_effort "high". User simulator gpt-5.2 (reasoning_effort low). Evaluation date 2026-03-02. Notes summary: 4 trial runs; seed 123.transcribed from table
Source excerptairline.pass_1=82.5Location In submissions/gemini-3-flash_sierra_2026-03-02/submission.json: results.airline.pass_1 (line 22 of file). model_name "Gemini 3 Flash". methodology.notes (line 56)
τ²-bench banking_knowledge (v1.0.1)High (default)High27.3%—Sierraτ²-bench leaderboard submission gemini-3-flash_sierra_2026-03-02 (sierra-research/tau2-bench, MIT)third-party · Retrieved 2026-11-13τ²-bench leaderboard submission gemini-3-flash_sierra_2026-03-02 (submission_type standard, tau2_bench_version 1.0.1). Submitter "Sierra", model_organization "Google": Run by Sierra, which operates τ²-bench (third-party measurement). Verified on the leaderboard (trajectories available, prompts not modified, no questions omitted). The value is pass^1 (%). reasoning_effort "high". User simulator gpt-5.2 (reasoning_effort low). Search configuration retrieval_config "terminal". Evaluation date 2026-03-02. Notes summary: rescored on 2026-07-15 following a scoring correction in v1.0.1; 4 trial runs; seed 123; 52 cases terminated due to max_steps and scored 0 points.transcribed from table
Source excerptbanking_knowledge.pass_1=27.32Location In submissions/gemini-3-flash_sierra_2026-03-02/submission.json: results.banking_knowledge.pass_1 (line 43 of file). model_name "Gemini 3 Flash". methodology.notes (line 56)
τ²-bench retail (v1.0.1)High (default)High76.8%—Sierraτ²-bench leaderboard submission gemini-3-flash_sierra_2026-03-02 (sierra-research/tau2-bench, MIT)third-party · Retrieved 2026-11-13τ²-bench leaderboard submission gemini-3-flash_sierra_2026-03-02 (submission_type standard, tau2_bench_version 1.0.1). Submitter "Sierra", model_organization "Google": Run by Sierra, which operates τ²-bench (third-party measurement). Verified on the leaderboard (trajectories available, prompts not modified, no questions omitted). The value is pass^1 (%). reasoning_effort "high". User simulator gpt-5.2 (reasoning_effort low). Evaluation date 2026-03-02. Notes summary: 4 trial runs; seed 123.transcribed from table
Source excerptretail.pass_1=76.75Location In submissions/gemini-3-flash_sierra_2026-03-02/submission.json: results.retail.pass_1 (line 29 of file). model_name "Gemini 3 Flash". methodology.notes (line 56)
τ²-bench telecom (v1.0.1)High (default)High91.2%—Sierraτ²-bench leaderboard submission gemini-3-flash_sierra_2026-03-02 (sierra-research/tau2-bench, MIT)third-party · Retrieved 2026-11-13τ²-bench leaderboard submission gemini-3-flash_sierra_2026-03-02 (submission_type standard, tau2_bench_version 1.0.1). Submitter "Sierra", model_organization "Google": Run by Sierra, which operates τ²-bench (third-party measurement). Verified on the leaderboard (trajectories available, prompts not modified, no questions omitted). The value is pass^1 (%). reasoning_effort "high". User simulator gpt-5.2 (reasoning_effort low). Evaluation date 2026-03-02. Notes summary: 4 trial runs; seed 123; 20 cases terminated due to too_many_errors and scored 0 points.transcribed from table
Source excerpttelecom.pass_1=91.23Location In submissions/gemini-3-flash_sierra_2026-03-02/submission.json: results.telecom.pass_1 (line 36 of file). model_name "Gemini 3 Flash". methodology.notes (line 56)

Related rankings

References