$ cat wiki/models/gemini-3-6-flash.md

Gemini 3.6 Flash

modelupdated 2026-09-13created 2026-07-22

Spec

Attribute	Value
Developer	Google DeepMind
Released	2026-07-21
Announced	2026-07-21
Context window	1,048,576 tokens (1M)
Pricing	$1.50/M input · $7.50/M output
License	proprietary
Availability	Gemini API, Google AI Studio, Android Studio, Vertex AI, consumer Gemini app, Google Search, GitHub Copilot

Benchmark	Gemini 3.6 Flash	Gemini 3.5 Flash
DeepSWE	49%	37%
SWE-Bench Pro	58.7%	55.1%
MLE Bench	63.9%	49.7%
GDPval-AA	1421	1349
OSWorld-Verified (Computer Use)	83%	78.4%
AA Intelligence Index	50	50
Time per task	~1.3 min	~2.7 min
Output tokens per task	~17% fewer	baseline
Output speed	304 t/s	unknown
Knowledge cutoff: March 2026 (up from January 2025 on 3.5 Flash).

Efficiency-over-intelligence design: The Artificial Analysis Intelligence Index is unchanged (50), but real agentic tasks complete in 1.3 min vs. 2.7 min — achieved by the model taking fewer tokens per step, not by reasoning better on each step. This makes it materially cheaper for agentic workloads.
Knowledge cutoff advance: March 2026 (15 months more recent than 3.5 Flash's January 2025 cutoff) — relevant for current-events queries in agentic search.
Lower output price: $7.50/M (down from $9.00/M) alongside the token-efficiency improvement — double benefit for agentic deployments where output cost dominates.
GitHub Copilot integration: Available in Copilot at launch (alongside Google's own surfaces).

Model	SWE-Bench Pro	Time/task	Price (in/out)	Notes
Gemini 3.6 Flash	58.7%	1.3 min	$1.50/$7.50	This model
Gemini 3.5 Flash	55.1%	2.7 min	$1.50/$9.00	predecessor
Claude Sonnet 5	63.2%	unknown	$2/$10	Anthropic mid-tier
Kimi K3	unknown	unknown	$0.30/$3/$15	Chinese MoE
DeepSeek V4	80.6% (SWE-bench Verified)	unknown	unknown	Open-weight SOTA

The published price disagrees with the catalogue's first-party endpoint, and the page's figure stands. spec-check run 90 (2026-09-12) reports input $1.5 vs $0.38; output $7.5 vs $1.88, both ~4.0× against Google (1차) (source).

The Spec row is not changed. Per CLAUDE.md, a page that cites a vendor announcement keeps the vendor's figure, and the cell above is faithful to the announcement this page cites. What was missing was the disclosure, which is what the schema asks for and what this entry supplies.

What is not established: which figure is correct, and what the catalogue's price is a price for — no tier, context band, modality or billing unit appears in the check's output. The ratio is an exact small-integer multiple in both cells at once, and it is on six of the eight conflicting pages, which is the shape of a tier or unit mismatch rather than of six independent errors — recorded as a pattern and not adopted as an explanation (source).

This has been reported by a red Action on every run since 2026-08-29 — 25 consecutive — and had reached no page until 2026-09-13. openrouter.ai is blocked from the daily run's sandbox, so the catalogue cannot be re-read here to settle it; that is a question for a run with egress.