Google's workhorse Flash-tier model, released 2026-07-21. Frontier-level quality tuned for token efficiency in coding, agentic execution, and multimodal work, with a 1M-token context window and thinking enabled.
Gemini 3.6 Flash is Google's workhorse Flash-tier model, released 21 July 2026 alongside Gemini 3.5 Flash-Lite. Google positions it around token efficiency rather than raw peak score: it beats its predecessor Gemini 3.5 Flash on every benchmark in Google's own model card while, per Google, using about 17% fewer output tokens to get there. It carries a 1,048,576-token input window, 65,536 output tokens, and thinking is on by default.
On TheRouter this model is served over the OpenAI-compatible chat completions API and is priced at Google's published Standard-tier rate: $1.50 per 1M input tokens, $7.50 per 1M output tokens (output includes thinking tokens), and $0.15 per 1M cached input tokens. Google additionally charges $1.00 per 1M tokens per hour to keep a cache alive; TheRouter's rate card carries the read rate only, so cache storage is not billed through TheRouter.
| Type | Rate |
|---|---|
| Input | $1.62 / 1M tokens |
| Output | $8.10 / 1M tokens |
| Release date | 2026-07-21deepmind.google ↗ | verified |
| Knowledge cutoff | March 2026deepmind.google ↗ | verified |
| Input token limit | 1,048,576ai.google.dev ↗ | verified |
| Output token limit | 65,536ai.google.dev ↗ | verified |
| Native modalities (upstream) | Input: text, image, video, audio, PDF. Output: text only. TheRouter exposes text, image, and PDF input with text output.ai.google.dev ↗ | verified |
| Supported capabilities | Thinking, function calling, structured outputs, code execution, context caching, Batch API, Google Search grounding, Google Maps grounding, file search. Computer use is in preview.ai.google.dev ↗ | verified |
| Not supported upstream | Audio generation, image generation, Live APIai.google.dev ↗ | verified |
| Vendor list price (Standard tier) | $1.50 / 1M input, $7.50 / 1M output, $0.15 / 1M cached input, plus $1.00 / 1M tokens per hour cache storage charged by Googleai.google.dev ↗ | verified |
| Output token efficiency | Google reports ~17% fewer output tokens than Gemini 3.5 Flash, attributed to the Artificial Analysis Indexdeepmind.google ↗ | to verify |
| Benchmark | Distribution | Score | Source |
|---|---|---|---|
SWE-Bench Pro Gemini 3.5 Flash 55.1%; Gemini 3.1 Pro 54.2% | 58.7%% | deepmind.google ↗ | |
Terminal-bench 2.1 Gemini 3.5 Flash 76.2%; Gemini 3.1 Pro 73.8% | 78.0%% | deepmind.google ↗ | |
DeepSWE v1.1 Gemini 3.5 Flash 37%; Gemini 3.1 Pro 12% | 49%% | deepmind.google ↗ | |
MLE-Bench Gemini 3.5 Flash 49.7%; Gemini 3.1 Pro 42.6% | 63.9%% | deepmind.google ↗ | |
OSWorld-Verified Gemini 3.5 Flash 78.4%; Gemini 3.1 Pro 76.2% | 83.0%% | deepmind.google ↗ | |
CharXiv (no tools) Highest in Google's comparison table | 85.2%% | deepmind.google ↗ | |
GDM-MRCR v2 (128k) Gemini 3.5 Flash 77.3%; Gemini 3.1 Pro 84.9% | 91.8%% | deepmind.google ↗ | |
GDM-MRCR v2 (1M) Gemini 3.5 Flash 26.6%; Gemini 3.1 Pro 26.3% | 54.0%% | deepmind.google ↗ | |
GDPval-AA v2 Gemini 3.5 Flash 1349; Gemini 3.1 Pro 965 | 1421Elo | deepmind.google ↗ |
Use the global api.therouter.ai endpoint shown below for new integrations; the legacy China accelerated endpoint is retired.
curl https://api.therouter.ai/v1/chat/completions -H "Content-Type: application/json" -H "Authorization: Bearer $THE_ROUTER_API_KEY" -d '{
"model": "google/gemini-3.6-flash",
"messages": [
{"role": "user", "content": "Summarize the key points from this input."}
]
}'Standard OpenAI-compatible chat completion. Thinking is enabled upstream by default and its tokens are billed as output.
curl https://api.therouter.ai/v1/chat/completions \
-H "Authorization: Bearer $THEROUTER_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model":"google/gemini-3.6-flash","messages":[{"role":"user","content":"Summarise this repo diff"}]}'Google released Gemini 3.6 Flash together with Gemini 3.5 Flash-Lite and a limited-access 3.5 Flash Cyber. The headline claim for 3.6 Flash is efficiency rather than a new peak: Google's model card shows it ahead of Gemini 3.5 Flash on every listed benchmark, and Google reports roughly 17% fewer output tokens for the same work, with the output rate dropping from $9.00 to $7.50 per 1M tokens.
It is the direct successor and leads on every benchmark Google published for the pair — for example 58.7% vs 55.1% on SWE-Bench Pro and 91.8% vs 77.3% on GDM-MRCR v2 at 128K. It is also cheaper on output ($7.50 vs $9.00 per 1M) and Google reports about 17% fewer output tokens for equivalent work.
No. Google's model page lists audio generation, image generation, and the Live API as unsupported. The model accepts multimodal input but returns text only. For image generation use a dedicated Gemini image model.
Thinking is enabled by default on this model and Google bills thinking tokens as output tokens. Google's own pricing page states that the $7.50 per 1M output rate includes thinking tokens, so usage figures include reasoning the model did not surface in the response.
| source | URL | retrieved | |
|---|---|---|---|
| Release date | deepmind.google ↗ | 2026-07-28 | verified |
| Knowledge cutoff | deepmind.google ↗ | 2026-07-28 | verified |
| Input token limit | ai.google.dev ↗ | 2026-07-28 | verified |
| Output token limit | ai.google.dev ↗ | 2026-07-28 | verified |
| Native modalities (upstream) | ai.google.dev ↗ | 2026-07-28 | verified |
| Supported capabilities | ai.google.dev ↗ | 2026-07-28 | verified |
| Not supported upstream | ai.google.dev ↗ | 2026-07-28 | verified |
| Vendor list price (Standard tier) | ai.google.dev ↗ | 2026-07-28 | verified |
| Output token efficiency | deepmind.google ↗ | 2026-07-28 | to verify |
| SWE-Bench Pro | deepmind.google ↗ | 2026-07-28 | verified |
| Terminal-bench 2.1 | deepmind.google ↗ | 2026-07-28 | verified |
| DeepSWE v1.1 | deepmind.google ↗ | 2026-07-28 | verified |
| MLE-Bench | deepmind.google ↗ | 2026-07-28 | verified |
| OSWorld-Verified | deepmind.google ↗ | 2026-07-28 | verified |
| CharXiv (no tools) | deepmind.google ↗ | 2026-07-28 | verified |
| GDM-MRCR v2 (128k) | deepmind.google ↗ | 2026-07-28 | verified |
| GDM-MRCR v2 (1M) | deepmind.google ↗ | 2026-07-28 | verified |
| GDPval-AA v2 | deepmind.google ↗ | 2026-07-28 | verified |
| Google ships Gemini 3.6 Flash and 3.5 Flash-Lite | blog.google ↗ | 2026-07-28 | verified |
| How does Gemini 3.6 Flash compare to Gemini 3.5 Flash? | deepmind.google ↗ | 2026-07-28 | to verify |
| Can it generate images or audio? | ai.google.dev ↗ | 2026-07-28 | to verify |
| Why is my completion_tokens count higher than the text I received? | ai.google.dev ↗ | 2026-07-28 | to verify |