Gemini 3.5 Pro Launch Remains Unconfirmed
NokiaPowerUser reported on August 23 that Gemini 3.5 Pro could launch that day, citing leaked specifications including a possible 2 million-token context window. Google has not announced a broad release: its July 21 blog post stated that Gemini 3.5 Pro was testing with partners and would become broadly available when ready.
A report from NokiaPowerUser on August 23 claimed that Google's Gemini 3.5 Pro could launch that day, but the report did not provide an official announcement or independently verifiable release details. The outlet described purported capabilities including advanced agentic coding, multi-step task execution, enhanced reasoning, and a context window of up to 2 million tokens.
Google's most recent directly relevant public statement points to a different status. In a July 21 post introducing Gemini 3.6 Flash, Gemini 3.5 Flash-Lite, and Gemini 3.5 Flash Cyber, Google wrote that Gemini 3.5 Pro was "currently testing with partners" and that it intended to make the model broadly available once ready. That post did not provide a public launch date, pricing, API availability, benchmark results, or context-window specification for Pro.
What Google has publicly released
Google introduced the Gemini 3.5 model family at Google I/O in May, beginning with Gemini 3.5 Flash. Google Cloud CEO Thomas Kurian described Flash as a model for agentic and coding workloads, and Google published benchmark figures for it including 76.2% on Terminal-Bench 2.1, 1656 Elo on GDPval-AA, and 83.6% on MCP Atlas.
Google's July post subsequently announced a newer Flash line:
- •Gemini 3.6 Flash, priced at $1.50 per million input tokens and $7.50 per million output tokens.
- •Gemini 3.5 Flash-Lite, which Google described as its fastest and most cost-effective 3.5-class model.
- •Gemini 3.5 Flash Cyber, used with the CodeMender code-security agent.
The same July post said Google had begun a pre-training run for Gemini 4. That statement establishes ongoing model development, but it does not verify the leaked Gemini 3.5 Pro specifications or an August 23 launch.
What remains unverified
The NokiaPowerUser report frames the alleged Pro model as a heavier-weight option for complex reasoning, full-stack coding, and autonomous software workflows. However, its claims about a 2 million-token context limit, architecture-level reasoning changes, and agent performance have not been corroborated by the Google materials provided.
For ML teams, the distinction matters because a public product launch normally requires concrete deployment information: supported modalities, API and console access, rate limits, token pricing, safety controls, tool-use interfaces, and reproducible evaluations. Until Google publishes those details, Gemini 3.5 Pro should be treated as a partner-tested model with unverified leak-based claims rather than an available production endpoint.
Across the wider model market, long-context and agentic claims require evaluation beyond a headline token limit. Teams comparing systems typically test retrieval accuracy over long inputs, tool-call reliability, latency under multi-step workloads, and total token consumption per completed task.
Key Points
- 1A reported August 23 launch remains unconfirmed, while Google's July update described Gemini 3.5 Pro as partner-tested rather than broadly available.
- 2The reported 2 million-token context window and advanced agent capabilities lack corroboration in the official Google materials provided.
- 3Long-context agent evaluations typically require retrieval, tool reliability, latency, and total-cost testing rather than token-limit comparisons alone.
Scoring Rationale
A broad Gemini 3.5 Pro release would be relevant to teams building coding and agentic systems, but the reported launch and specifications remain unverified. Google's official material confirms partner testing, not public availability, limiting the immediate operational impact.
Sources
Primary source and supporting public references used for this report.
Practice with real Ad Tech data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all Ad Tech problems
