Model
Model explorer

Gemini 3.6 Flash

CLOSED
Google · Gemini 3.6 family · released Jul 21, 2026

Google DeepMind's new workhorse Flash model, released 2026-07-21 alongside Gemini 3.5 Flash-Lite and the gated Gemini 3.5 Flash Cyber (there was NO Pro release in this wave -- the flagship stayed on Gemini 3.1 Pro; Gemini 3.5 Pro remained 'coming soon'/in partner testing). Built on Gemini 3.5 Flash, it is pitched as a straight quality-and-efficiency upgrade: better coding, knowledge work and multimodal performance while using ~17% fewer output tokens for the same work, and it CUTS the output price from $9.00 to $7.50/M (input unchanged at $1.50/M; cache-hit input ~$0.15/M). 1M-token input / 64K output context, natively multimodal (text/image/audio/video in, text out), knowledge cutoff March 2026. Model ID gemini-3.6-flash, generally available on the Gemini API, AI Studio, Vertex, the Gemini apps and Antigravity. All benchmark figures here are Google self-reported from the official model card -- no independent per-benchmark scores exist yet (Artificial Analysis publishes only its composite Intelligence Index of 50 for this model, no per-benchmark numerals). Re-check trackers after independent evals land.

ReasoningCodingVisionFunction callingTool useAgentic
2630.5
Elo · rank #10
Parameters
Undisclosed
Active params
Undisclosed
Context
1M tokens
Architecture
Sparse Mixture-of-Experts Transformer (configuration undisclosed); natively multimodal reasoning model in the Gemini 3 series
License
Proprietary
Languages
API price (in/out)
$1.5 / $7.5
Modalities
text · vision · audio · video
Benchmark results
Bar shows position within the tracked field; marker = field best
GDPval-AAAgents1421#18
best: Claude Fable 5 · 1932
OSWorld-VerifiedAgents83.0%#3
best: Claude Fable 5 · 85.0%
SWE-bench ProCoding58.7%#12
best: Claude Fable 5 · 80.0%
Terminal-Bench 2.0Coding78.0%#10
best: GPT-5.6 · 88.8%
Run it locally
Closed weights — available via API only. No local deployment.
Input / M tok
$1.5
Output / M tok
$7.5
Gemini 3.6 family
Elo progression across releases
API price $1.5/$7.5 · each benchmark row carries its own source badge (see methodology)