Google Launches Gemini 3.6 Flash with 17% Output Token Efficiency Gain
Google debuts three new Gemini models including 3.6 Flash with improved efficiency, while 3.5 Pro remains in limited testing.
Gemini 3.6 Flash: Efficiency and Performance Gains
Google has released Gemini 3.6 Flash, delivering a 17% reduction in output token usage compared to Gemini 3.5 Flash according to the Artificial Analysis Index. The model is priced at $1.50 per million input tokens and $7.50 per million output tokens.
Across multiple benchmarks, Gemini 3.6 Flash demonstrates meaningful improvements:
- DeepSWE: 49% versus 37% on Gemini 3.5 Flash
- OSWorld-Verified: 83.0% versus 78.4% on Gemini 3.5 Flash
- GDPval-AA v2: 1421 versus 1349 on Gemini 3.5 Flash
- MLE Bench: 63.9% versus 49.7% on Gemini 3.5 Flash
Gemini 3.5 Flash-Lite: Speed and Cost Optimized
Google has also introduced Gemini 3.5 Flash-Lite, a lightweight model delivering 350 output tokens per second according to the Artificial Analysis Index. It is priced at $0.3 per million input tokens and $2.5 per million output tokens.
The model shows substantial gains over Gemini 3.1 Flash-Lite:
- Terminal-Bench 2.1: 54% versus 31%
- GDM-MRCR v2: 72.2% versus 60.1%
- GDPval-AA v2: 1140 versus 642
Compared to Gemini 3 Flash, Gemini 3.5 Flash-Lite achieves:
- SWE-Bench Pro: 54.2% versus 49.6%
- OSWorld-Verified: 74.0% versus 65.1%
Specialized and Forthcoming Models
Gemini 3.5 Flash Cyber will be exclusively available to governments and trusted partners as part of a limited-access pilot program.
Gemini 3.5 Pro is currently testing with partners, with Google planning to make it broadly available as soon as it’s ready.
Google has started its most ambitious pre-training run yet for Gemini 4.
Source: Google Official Blog
Developments since publication
-
Tencent released one new AI model on August 28, 2026: Hy4 preview. Source
-
Alibaba released Qwen3.8 Flash on August 26, 2026. Source
-
Zhipu AI released GLM-5.3 Flash on August 26, 2026. Source
-
GLM-5.3-Flash is described as the first natively multimodal GLM-5 model. Source
-
GLM-5.3-Flash has a 320B/18B parameter configuration and a 1M token context window. Source
-
GLM-5.3-Flash is released under an MIT licence (open weights). Source
-
GLM-5.3-Flash is listed at $0.15 per 1M input tokens and $0.50 per 1M output tokens. Source
-
Qwen3.8-Flash-Next is an open-weight preview of the Qwen4 architecture. Source
-
Qwen3.8-Flash-Next has 6 billion active parameters and a 262K native context window. Source
-
DeepSeek released DeepSeek-V4-Flash-Vision-Exp on August 21, 2026. Source
-
Google released Gemini 3.5 Transcribe as two separate endpoints — a streaming endpoint and a non-streaming endpoint. Source
-
Gemini 3.5 Transcribe's streaming endpoint delivers sub-second transcription but drops speaker diarisation. Source
-
An article published on August 28, 2026 claims OpenAI is on track to reach its internal AGI bar by end of 2026. Source
-
Alibaba's WAN 3.0 video model generates up to 30 seconds of 1080p video with audio in a single pass. Source
-
WAN 3.0 is listed on fal at $0.05/$0.10/$0.20 per second (standard), and $0.068/$0.14/$0.28 per second for the Prime (accelerated) SKU. Source
-
Gemini 3.7 Flash was released by Google on August 13, 2026. Source
-
Qwen Image 3.0 Pro was released by Alibaba on August 5, 2026. Source
-
Qwen Image 3.0 was released by Alibaba on August 5, 2026. Source
-
The most recent frontier AI model release tracked by AI Release Tracker is Muse Spark 1.2 by Meta, released on August 5, 2026. Source
-
Qwen3.8-Max was released by the Qwen team on August 3, 2026. Source
-
DeepSeek-V4-Flash-0731 was released by DeepSeek on July 31, 2026. Source
-
Claude Opus 5 was released by Anthropic on July 24, 2026. Source
-
Google released three models on July 21, 2026: Gemini 3.6 Flash, Gemini 3.5 Flash-Lite, and Gemini 3.5 Flash Cyber. Source
-
Kimi K3 was released by Moonshot AI on July 16, 2026. Source
-
Kimi K3 was launched at the World AI Conference in Shanghai. Source
-
Kimi K3 has a 1 million token context window. Source
-
Kimi K3 supports native multimodal input. Source
-
Kimi K3 is described as the largest open-weight model ever released at the time of its launch. Source
-
Grok 4.5 was released by SpaceXAI (xAI) on July 8, 2026. Source
-
OpenAI released three GPT-5.6 variants — Sol, Terra, and Luna — on June 26, 2026. Source
-
Meta Muse Spark 1.1 was released on July 9, 2026. Source
Irish pronunciation
All FoxxeLabs components are named in Irish. Click ▶ to hear each name spoken by a native Irish voice.