Google Ships Gemini 3.7 Flash as Gemini 3.5 Pro Remains Missing

Google released Gemini 3.7 Flash on August 13, 2026, describing it as its “most intelligent workhorse model yet for coding and agents.” The launch came three weeks after Gemini 3.6 Flash became generally available on July 21, and it extended a rapid cadence in the Flash line while the previously announced flagship, Gemini 3.5 Pro, remained unreleased. For developers and enterprise buyers, the sequence signaled a clear priority: shipping frequent, incremental improvements in cost-efficient Flash models over delivering a higher-end Pro model first promised for June 2026.

The timing matters because it reveals how Google is positioning its AI portfolio amid intensifying competition and growing enterprise demand for reliable, deployable tools. Rather than waiting for a headline-grabbing frontier model, the company is shipping practical updates aimed at software engineering, web development, and agentic workflows. The absence of Gemini 3.5 Pro—despite public commitments from CEO Sundar Pichai—also raises questions about the difficulty of frontier model development and whether Google is deliberately shifting resources toward models that can be integrated into products and agent platforms more quickly.

A new workhorse arrives on an accelerated schedule

Gemini 3.7 Flash’s release on August 13, 2026 followed the July 21 general availability of Gemini 3.6 Flash by only a few weeks. That cadence marks a departure from the slower, more infrequent model launches that characterized earlier generations. Google described the new model as its most intelligent workhorse model yet for coding and agents, and it is now the default model for Google’s Antigravity agent, according to the company’s AI blog. The default status means developers using Antigravity will interact with Gemini 3.7 Flash unless they actively choose another model, embedding the update directly into Google’s agent product.

The model delivers what Google called substantial improvements in software engineering, web development, and agentic workflows. It is available at an introductory price through December 31, 2026, though the company did not confirm exact pricing in official sources. A prior model, Gemini 3.6 Flash, was reported by Mashable in June 2026 to be priced lower than its predecessor, suggesting that Google is continuing to push cost efficiency across the Flash series. The introductory pricing window also points to a deliberate effort to attract developers during a critical adoption period.

Koray Kavukcuoglu, CTO of Google DeepMind and Chief AI Architect at Google, highlighted the model’s capabilities in coding and agent execution. The emphasis on agentic workflows is notable: Google is positioning Flash not just as a general-purpose language model but as a component of automated systems that can perform multi-step tasks. That aligns with the company’s broader push into agent-based products, where models must handle tool use, planning, and execution with low latency and predictable costs.

The flagship that didn’t ship

While the Flash line accelerated, Gemini 3.5 Pro remained absent. Google CEO Sundar Pichai initially stated the model would launch in June 2026, then later suggested July 17, 2026, according to Mashable and Tech Insider. Both dates passed without a release. As of early September 2026, Gemini 3.5 Pro had still not been released. Mashable reported in July 2026 that the model was “still MIA.” At Google I/O 2026, Pichai had announced Gemini 3.5 Flash and discussed the delayed rollout of the Pro model, but the Pro tier has yet to appear.

The delay has not been fully explained. Google has not published official benchmarks comparing Gemini 3.7 Flash with the unreleased Gemini 3.5 Pro, leaving the performance relationship between the two models unclear. The absence of a flagship Pro model after repeated public timelines has led to speculation about internal development challenges or a strategic reprioritization toward Flash variants. Without official data, developers cannot determine whether the new Flash model now covers tasks that would previously have required a Pro-tier model.

For observers, the situation is unusual. Google has historically used its Pro tier to showcase frontier capabilities, while Flash models have been positioned as faster and cheaper alternatives. By shipping multiple Flash generations without a corresponding Pro release, Google may be signaling that the traditional tier structure is becoming less relevant to its commercial strategy. The company appears more focused on models that can be deployed immediately in products like Antigravity than on maintaining a separate flagship for benchmark leadership.

What Google is signaling to developers and enterprises

Google’s messaging around Gemini 3.7 Flash centers on cost efficiency and high performance for developers and enterprises. The model’s improvements in software engineering and web development are aimed at practical workloads, not just benchmark scores. Making it the default model for the Antigravity agent also ties the release directly to Google’s agentic product strategy, where reliability and integration matter more than raw capability in isolation.

“Most intelligent workhorse model yet for coding and agents” — Google AI Blog, August 2026

This focus reflects broader market demand. Enterprises increasingly need AI models that can be deployed reliably, at scale, and with predictable costs. Flash variants are optimized for speed, cost, and agent workflows, which makes them attractive for production environments where latency and operating expenses matter as much as raw capability. The introductory pricing through December 31, 2026 also suggests a push to attract developers during a critical adoption window. While exact pricing was not confirmed, the pattern of lower-cost Flash releases indicates that Google is competing on total cost of ownership as much as on model quality.

The unresolved benchmark question and what comes next

One of the biggest open questions is how Gemini 3.7 Flash compares with the unreleased Gemini 3.5 Pro. Without official benchmarks, developers cannot easily assess whether the Flash model is now sufficient for tasks that would previously have required a Pro-tier model. Google has not clarified whether the delay of 3.5 Pro means the model will be released later, replaced by a different flagship, or absorbed into the Flash roadmap. The lack of clarity leaves enterprise planning in a difficult position, as buyers must decide whether to build on the Flash line or wait for a Pro model that may or may not arrive.

The rapid iteration from Gemini 3.6 Flash to 3.7 Flash suggests Google is adopting a more continuous release model, at least for its workhorse tier. That approach can deliver incremental value quickly, but it also creates challenges for enterprise planning, as customers must decide whether to adopt each new version or wait for the next one. Frequent updates can improve performance and fix issues, but they also require integration work and can complicate long-term procurement decisions.

The delay of the Pro model may indicate that frontier model development is becoming more complex, requiring longer training runs, more rigorous safety testing, or greater compute resources. Alternatively, Google may be prioritizing models that can be integrated into products like Antigravity immediately, rather than pursuing a separate flagship that may not translate into near-term revenue. Both explanations are consistent with the available evidence, and Google has not publicly clarified its reasoning.

Looking ahead, the contrast between the fast-moving Flash line and the stalled Pro release will likely define Google’s AI narrative in the coming months. If Gemini 3.5 Pro eventually ships, it will face a market that has already been conditioned to expect frequent, practical updates from the Flash series. If it does not, Google’s decision to emphasize deployable, cost-efficient models may mark a lasting shift in how it competes in the enterprise AI market. Either way, developers and enterprise buyers will be watching closely for benchmarks, pricing details, and any official word on the missing flagship.