Google has released Gemini 3.6 Flash, and Gemini 4 has also started pre-training

According to DynaBeat monitoring, Google has released Gemini 3.6 Flash, focusing on programming, multimodality, and multi-step Agent workflows. The new model reduces inference steps, tool invocations, and execution loops.

In the Artificial Analysis Index, its output Token usage is 17% lower than Gemini 3.5 Flash. The API input price remains at $1.5 per million Tokens, while the output price has decreased from $9 to $7.5.

Performance has also improved. DeepSWE has increased from 37% to 49%, MLE Bench from 49.7% to 63.9%, and OSWorld-Verified from 78.4% to 83%.

The model is now officially open, supporting contexts of up to 1 million Tokens and a maximum output of 64,000 Tokens. Its most significant change is to enable the Agent to take shorter routes, spend fewer Tokens, and lower the cost of long tasks.

Google has also teased that Gemini 3.5 Pro is undergoing partner testing and will be fully launched upon completion.

Gemini 4 has also commenced pre-training, marking Google's most ambitious pre-training round to date. However, the official has not disclosed a release date or more details.