GEMINI 3.6 FLASH — KEY FACTS (LAUNCHED JULY 21, 2026)
● Model ID: gemini-3.6-flash (gemini-3.6-flash-tiered was an internal preview label)
● Price: $1.50/M input · $7.50/M output (output down from $9/M on Gemini 3.5 Flash)
● Batch / Flex pricing: $0.75/$3.75/M (50% off standard)
● Context window: 1,048,576 input tokens · 65,536 output tokens
● Knowledge cutoff: March 2026 (up from January 2025 on 3.5 Flash — 14-month jump)
● Speed: ~304 tokens/sec — among the fastest in its price tier
● AA Intelligence Index: 50 — identical to Gemini 3.5 Flash
● Average task time (AA): 1.3 min (down from 2.7 min on 3.5 Flash)
● Output token reduction: ~17% fewer tokens on AA Index vs 3.5 Flash
● Multimodal input: Text, image, audio, video, PDF — text output only
● Availability: Google AI Studio, Gemini API, Gemini app, Android Studio, Vertex AI, GitHub Copilot
● Cache pricing: $0.15/M cached input tokens
The One Finding That Defines This Model
Gemini 3.6 Flash is faster and cheaper than Gemini 3.5 Flash. It is not measurably smarter on independent benchmarks. As Artificial Analysis confirmed, both Gemini 3.5 Flash and Gemini 3.6 Flash score 50 on the AA Intelligence Index — a composite of nine evaluations including Terminal-Bench v2.1, GPQA Diamond, SWE-bench, and HLE. The score did not move. As HackerNoon's analysis explains, this does not mean the model did not improve — composite indices can stay flat while individual capabilities move. Google's own benchmarks show real gains on coding agents (DeepSWE 49% vs 37%), computer use (OSWorld-Verified 83.0% vs 78.4%), and ML research (MLE-Bench 63.9% vs 49.7%). Those gains did not move the aggregate integer score.
The practical conclusion from Memeburn's benchmark guide: "Gemini 3.6 Flash improves the economics of workloads already suited to the Flash tier. It does not establish that Google has closed the capability gap at the upper end of complex reasoning and coding." It is a faster, cheaper 3.5 Flash — not a new capability tier.
Benchmark Results — Google vs Independent
| Benchmark | Gemini 3.5 Flash | Gemini 3.6 Flash | Source |
|---|
| AA Intelligence Index | 50 | 50 (unchanged) | Artificial Analysis (independent) |
| DeepSWE | 37% | 49% | Google (vendor benchmark) |
| OSWorld-Verified | 78.4% | 83.0% | Google (vendor benchmark) |
| MLE-Bench | 49.7% | 63.9% | Google (vendor benchmark) |
| SWE-Bench Pro | 55.1% | 58.7% | Google (vendor benchmark) |
| Avg task time (AA) | 2.7 min | 1.3 min | Artificial Analysis (independent) |
| Output tokens | Baseline | ~17% fewer | Artificial Analysis (independent) |
Google vendor benchmarks are self-selected and self-run. Artificial Analysis numbers are independent. Both are valid — they answer different questions. The unchanged AA Intelligence Index score of 50 is the most important single number for teams deciding whether to migrate from 3.5 Flash to 3.6 Flash.
Pricing — Where It Sits in the Market
| Model | Input /1M | Output /1M | AA Index |
|---|
| GPT-5.6 Luna (OpenAI) | $0.20 | $1.20 | Higher than 3.6 Flash |
| Gemini 3.6 Flash | $1.50 | $7.50 | 50 |
| Gemini 3.5 Flash (prev) | $1.50 | $9.00 | 50 |
| Claude Sonnet 5 | $3.00 (from Sept 1) | $15.00 | Higher than 3.6 Flash |
| Gemini 3.1 Pro | $2.00 | $12.00 | 46 (below 3.6 Flash) |
As BenchLM's August 2026 pricing comparison notes, GPT-5.6 Luna is cheaper than Gemini 3.6 Flash on both input ($0.20 vs $1.50) and output ($1.20 vs $7.50) — and scores higher on independent benchmarks. For teams not locked into the Google ecosystem, Luna is the stronger cost-performance case at this capability tier. The case for Gemini 3.6 Flash is Google-native: Search grounding, Google Workspace integration, Vertex AI deployment, and the 304 tokens/sec speed that makes it one of the fastest models in any price class.
The Knowledge Cutoff Jump — Most Underrated Upgrade
According to DataNorth AI's launch coverage, the knowledge cutoff moved from January 2025 to March 2026 — a 14-month jump. For applications that ask Gemini about recent AI models, current events, or 2025-2026 developments, this is the most practically significant upgrade. Gemini 3.5 Flash's January 2025 cutoff meant it had no knowledge of GPT-4.5, Claude 3.7, the 2025 AI model releases, or any development from the most active 18 months in AI history. 3.6 Flash covers through March 2026.
Should You Migrate From Gemini 3.5 Flash?
Yes — migrate if:
You are on Gemini 3.5 Flash. The input price is the same ($1.50/M), the output price is lower ($7.50 vs $9/M), average task time halved, and output tokens reduced 17%. The 14-month knowledge cutoff jump is an immediate practical improvement. Per CometAPI's migration guide, most production workloads see roughly 12-23% cost reduction at equal token usage. The model ID change is the only required update: replace gemini-3.5-flash with gemini-3.6-flash.
Evaluate alternatives if:
You are not Google-ecosystem-committed. GPT-5.6 Luna at $0.20/$1.20/M is dramatically cheaper with a higher AA Index score. DeepSeek V4 Flash at $0.14/$0.28/M leads on Terminal-Bench for agentic coding. Claude Sonnet 5 (through August 31 at $2/M) has a higher intelligence ceiling. The case for Gemini 3.6 Flash outside the Google ecosystem is speed and multimodal input — not price or intelligence leadership.
Sources: Kie.ai launch coverage · Artificial Analysis benchmark data · Memeburn benchmark guide · HackerNoon intelligence analysis · BenchLM pricing comparison · CometAPI migration guide · Related: Gemini 3.6 Flash vs Claude Sonnet 5 → · Gemini 3.6 vs 3.5 Flash vs GPT-5.6 Luna →