LIVE
Third-party cyber evaluations involving OpenAI models04/08/26 · OpenAI|Video-DeepResearch: Towards the Next-Generation Multimodal Deepresearch Agent04/08/26 · OpenAI|Mistral's Shieldstral: 3B open-weights model for multimodal moderation04/08/26 · Mistral AI|NVIDIA Joins NSF State and Regional AI Hubs Program to Expand AI Research and Education Across the US04/08/26 · NVIDIA|Apple says more ex-employees may have taken confidential data to OpenAI04/08/26 · OpenAI|NVIDIA Alpamayo 2 Super, the Frontier Open Model for Robotaxis and Autonomous Vehicles, Now Available for Commercial Use04/08/26 · NVIDIA|As AI Increases Demands on Memory, Storage Steps Up04/08/26 · NVIDIA|Deploy local agents everywhere with LFM2.5-2.6B04/08/26 · Hugging Face|AI Leaders Propose SAFE Guidelines for Cybersecurity Transparency04/08/26 · NVIDIA|AI-Generated Images Discourage Me from Reading Your Blog04/08/26|Disrupting a Criminal Scam Operation04/08/26 · OpenAI|Mariano-Florentino (Tino) Cuéllar to join Anthropic as Chief Global Affairs Officer04/08/26 · Anthropic|Third-party cyber evaluations involving OpenAI models04/08/26 · OpenAI|Video-DeepResearch: Towards the Next-Generation Multimodal Deepresearch Agent04/08/26 · OpenAI|Mistral's Shieldstral: 3B open-weights model for multimodal moderation04/08/26 · Mistral AI|NVIDIA Joins NSF State and Regional AI Hubs Program to Expand AI Research and Education Across the US04/08/26 · NVIDIA|Apple says more ex-employees may have taken confidential data to OpenAI04/08/26 · OpenAI|NVIDIA Alpamayo 2 Super, the Frontier Open Model for Robotaxis and Autonomous Vehicles, Now Available for Commercial Use04/08/26 · NVIDIA|As AI Increases Demands on Memory, Storage Steps Up04/08/26 · NVIDIA|Deploy local agents everywhere with LFM2.5-2.6B04/08/26 · Hugging Face|AI Leaders Propose SAFE Guidelines for Cybersecurity Transparency04/08/26 · NVIDIA|AI-Generated Images Discourage Me from Reading Your Blog04/08/26|Disrupting a Criminal Scam Operation04/08/26 · OpenAI|Mariano-Florentino (Tino) Cuéllar to join Anthropic as Chief Global Affairs Officer04/08/26 · Anthropic|
🇺🇸 United States visionReleased July 2026

Gemini 3.6 Flash

Google
Verdict

A versatile model, balanced across most use cases.

Overview

Gemini 3.6 Flash, released July 21 2026, becomes the default model in the Gemini family, succeeding 3.5 Flash. 1,048,576-token context, 65,536 max output, multimodal input (text, image, video, audio, PDF). March 2026 knowledge cutoff versus January 2025 for its predecessor — a 14-month jump. Unusually, output pricing dropped ($7.50 from $9) while the model uses roughly 17% fewer output tokens, compounding the savings. It leads on long context (91.8% on GDM-MRCR v2, best in its comparison set) and computer use (83% on OSWorld-Verified), but trails GPT-5.6 Luna and Grok 4.5 on coding. Around 280 tokens per second.

Skill profile

Arena Elo1483

Not disclosed

Strengths

  • World-class on Arena
  • Very long context

Limitations

  • No native GDPR guarantee

Who is it for

A good fit if…
    Skip it if…
    • you handle sensitive EU data

    Access & availability

    Paid API per token

    Key specifications

    Context
    1M
    Input price
    $1.50 $/M
    Output price
    $7.50 $/M
    Speed
    Price not auditedBenchmarks not audited

    Estimate your monthly cost

    Per-token API pricing
    $98/ month with Gemini 3.6 Flash

    Indicative estimate based on standard API rates (excluding caching, batch and volume discounts). Always check the official pricing before committing.

    Privacy

    Not GDPR-compliant
    Advanced data · for experts
    Architecture, modalities, detailed cost, full benchmarks
    Architecture
    Size
    Cutoff date
    Inputs
    Outputs
    License
    commercial
    Hosting
    Hallucination score
    Estimated cost (API)
    Typical exchange (~3k in / 1k out)
    ≈ $0.0120
    1M in + 1M out
    ≈ $9.00
    All benchmarks
    Arena Elo1483
    MMLU
    GPQA
    HumanEval
    SWE-Bench
    MATH