Model Tracker
Gemini 3.5 Flash
Gemini 3.5 Flash is an AI model made by Google. Google released it on 19 May 2026. It takes text, image, video, audio and returns text.
En anglais
Gemini 3.5 Flash is an AI model made by Google. Google released it on 19 May 2026. It takes text, image, video, audio and returns text. Its context window is 1,048,576 tokens. Its longest answer is 65,536 tokens. Through the API it costs $1.50 per 1M tokens for input and $9.00 per 1M tokens for output. Its knowledge cutoff is January 2025.
- Released
- 19 May 2026 — ai.google.dev, read 16 October 2026
- Status
- Not confirmed: two readings disagree: deprecated and ga
- Takes
- Text, Image, Video, Audio — deepmind.google, read 16 October 2026
- Returns
- Text — deepmind.google, read 16 October 2026
- Open weights
- No — models.dev, read 20 November 2026, from public catalogs that agree (models.dev, Epoch AI), not yet checked on the maker’s page
- Licence
- Not confirmed: The official model card and documentation do not name a licence for Gemini 3.5 Flash weights.
- Context window
- 1,048,576 tokens — models.dev, read 20 November 2026, from public catalogs that agree (models.dev, OpenRouter), not yet checked on the maker’s page
- Longest answer
- 65,536 tokens — ai.google.dev, read 16 October 2026
- Knowledge cutoff
- January 2025 — ai.google.dev, read 16 October 2026
- API price, input
- $1.50 per 1M tokens — models.dev, read 20 November 2026, from public catalogs that agree (models.dev, OpenRouter, LiteLLM), not yet checked on the maker’s page
- API price, output
- $9.00 per 1M tokens — models.dev, read 20 November 2026, from public catalogs that agree (models.dev, OpenRouter, LiteLLM), not yet checked on the maker’s page
- Where
- Not confirmed: two readings disagree: api, app, cloud and api, app
- Parameters
- Not confirmed: The official model card and documentation do not publish a parameter count for Gemini 3.5 Flash.
- Model card
- https://deepmind.google/models/model-cards/gemini-3-5-flash/ — deepmind.google, read 16 October 2026
Some figures come from public model catalogs that agree; each is being checked on the maker’s page.
In the news
- Google Research Releases Harness² for Agents That Rewrite Their Setup (17 October 2026)
Benchmarks
Measured by others, not by the maker: each score with the name the measurer gives the model and the setting it ran, its day and its licence.
Arena
- Text Arena, Gemini 3.5 Flash (High): 1,479 (1,475–1,482, 47,470 votes), published 19 November 2026
- Text Arena, Gemini 3.5 Flash (Medium): 1,476 (1,472–1,480, 46,163 votes), published 19 November 2026
- Code Arena (WebDev), Gemini 3.5 Flash (High): 1,500 (1,492–1,507, 9,122 votes), published 20 November 2026
- Code Arena (WebDev), Gemini 3.5 Flash (Medium): 1,490 (1,483–1,497, 11,842 votes), published 20 November 2026
- Vision Arena, Gemini 3.5 Flash (Medium): 1,282 (1,276–1,289, 15,320 votes), published 19 November 2026
- Vision Arena, Gemini 3.5 Flash (High): 1,281 (1,275–1,288, 15,228 votes), published 19 November 2026
Arena’s leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0: ratings from people’s votes between two anonymous models, as Arena publishes them. Data · CC BY 4.0
Epoch AI
- Epoch Capabilities Index, Gemini 3.5 Flash: 155.3 (153.5–157.3), as of 21 November 2026
- GPQA Diamond, gemini-3.5-flash_high: 92.8% (± 1.6), run 22 May 2026
- GPQA Diamond, gemini-3.5-flash_low: 88.9% (± 2.2), run 6 August 2026
- GPQA Diamond, gemini-3.5-flash_minimal: 86.4% (± 2.4), run 15 July 2026
- FrontierMath, tiers 1–3, gemini-3.5-flash_high: 62.8% (± 2.9), run 10 June 2026
- FrontierMath, tier 4, gemini-3.5-flash_high: 26.8% (± 7.0), run 10 June 2026
- SWE-bench Verified, gemini-3.5-flash_high: 79.3% (± 1.8), run 1 June 2026
- OTIS Mock AIME 2024–2025, gemini-3.5-flash_high: 95.6% (± 2.7), run 25 May 2026
- OTIS Mock AIME 2024–2025, gemini-3.5-flash_low: 88.9% (± 4.7), run 6 August 2026
- OTIS Mock AIME 2024–2025, gemini-3.5-flash_minimal: 80.0% (± 6.0), run 15 July 2026
- SimpleQA Verified, gemini-3.5-flash_high: 66.2% (± 1.5), run 27 August 2026
Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks, CC BY 4.0: Epoch AI’s own runs and its Capabilities Index. Data · CC BY 4.0
Questions
When was Gemini 3.5 Flash released?
Google released Gemini 3.5 Flash on 19 May 2026.
What is the context window of Gemini 3.5 Flash?
The context window of Gemini 3.5 Flash is 1,048,576 tokens, as Google prints it.
How much does the Gemini 3.5 Flash API cost?
On the standard tier, Gemini 3.5 Flash costs $1.50 per 1M tokens for input and $9.00 per 1M tokens for output, as Google lists it.
Are the weights of Gemini 3.5 Flash open?
No: Google does not publish the weights of Gemini 3.5 Flash.
What does Gemini 3.5 Flash take and return?
Gemini 3.5 Flash takes text, image, video, audio and returns text.
What is the knowledge cutoff of Gemini 3.5 Flash?
The knowledge cutoff of Gemini 3.5 Flash is January 2025.
More from Google
- CodeGemma
- DiffusionGemma
- EmbeddingGemma 2
- Gemini 1.0 Pro Vision
- Gemini 1.5 Flash
- Gemini 1.5 Flash-8B
- Gemini 1.5 Pro
- Gemini 2.0 Flash
- Gemini 2.0 Flash Experimental
- Gemini 2.0 Flash-Lite
- Gemini 2.5 Computer Use
- Gemini 2.5 Deep Think
- Gemini 2.5 Flash
- Gemini 2.5 Flash Image
- Gemini 2.5 Flash Live
- Gemini 2.5 Flash TTS
- Gemini 2.5 Flash-Lite
- Gemini 2.5 Pro
- Gemini 2.5 Pro TTS
- Gemini 3 Flash
- Gemini 3 Pro
- Gemini 3.1 Flash Live
- Gemini 3.1 Flash TTS
- Gemini 3.1 Flash-Lite
Sources
- Gemini 3.5 Flash - Model Card (read 16 October 2026)
- Release notes (read 16 October 2026)
- Gemini 3.5 Flash (read 16 October 2026)
- Gemini Developer API pricing (read 16 October 2026)
- What's new in Gemini 3.5 Flash (read 16 October 2026)
- models.dev model catalog (read 20 November 2026)
- OpenRouter model catalog (read 20 November 2026)
- LiteLLM model catalog (read 20 November 2026)
- Epoch AI model catalog (read 20 November 2026)