Gemini 3.5 Pro API Release Watch: What Google Has Confirmed
TL;DR
- Google confirmed Gemini 3.5 Pro exists and is being used internally.
- It is not listed in the public Gemini API model catalog as of July 19, 2026.
- Google has not published a public Gemini 3.5 Pro model ID, API price, context limit, rate limit, or GA/preview status.
- Reuters reported on July 16 that the launch had been delayed after the model fell short of internal performance goals, particularly in coding. This is attributed reporting, not an official Google technical specification.
- Gemini 3.5 Flash is already generally available. Do not describe the Flash and Pro releases as being in the same state.
- Use available routes now and keep model selection configurable. Update production plans only after the exact Pro route is documented and verified.
Current Gemini 3.5 Pro API status
The table below separates official information from reported information and unknowns.
| Item | Status on July 19, 2026 | Evidence level |
|---|---|---|
| Gemini 3.5 Pro exists | Confirmed; Google says it is used internally | Official Google announcement |
| Public Gemini API availability | Not publicly listed | Official Gemini API model catalog |
| Public model ID | Not confirmed | No public API listing |
| API pricing | Not confirmed | No official pricing row for Gemini 3.5 Pro |
| GA, preview, or waitlist status | Not publicly specified for the Gemini API | Official status not published |
| Wider rollout timing | Google's May target has passed without a public API listing | Official announcement plus current docs check |
| Reported delay | Reuters, citing Bloomberg, reported a further delay tied to internal performance goals | Attributed media reporting |
| EvoLink route | Not available until upstream access and EvoLink verification are complete | EvoLink product status |
What Google officially confirmed at I/O 2026
In its May 19 Gemini 3.5 announcement, Google launched Gemini 3.5 Flash and said it was also working on Gemini 3.5 Pro. Google stated that Pro was already being used internally and would be rolled out more broadly the following month.
That statement confirms the model's development, but it does not confirm a public API launch. As of July 19, the public Gemini API model catalog lists Gemini 3.5 Flash as stable and Gemini 3.1 Pro as a preview model; it does not list Gemini 3.5 Pro.
For a developer-facing release watch, the public API listing is the operative checkpoint. A product mention, internal test, consumer experiment, or media report is not the same as a documented API route.
What the delay reporting says—and what it does not prove
Reuters reported on July 16, citing Bloomberg News, that Google had delayed Gemini 3.5 Pro because the model had not met internal performance goals, particularly for coding. That report explains the current news cycle and developer concern, but it should remain explicitly attributed.
What remains unconfirmed
Do not use the following as facts until Google publishes them in official model documentation or release notes:
- the exact Gemini 3.5 Pro API model string;
- Gemini API versus Vertex AI channel timing;
- preview, GA, allowlist, or regional rollout status;
- input and output token pricing;
- context window and maximum output;
- function calling, structured output, grounding, caching, or background execution support;
- quotas, rate limits, regions, and service tiers;
- benchmark results, production reliability, and cost per successful task.
Rumored values should not appear in the main facts table, title, metadata, or production recommendation.
What EvoLink users can use now
Waiting for Gemini 3.5 Pro does not require pausing model evaluation.
| Need now | Available EvoLink path | Why it is the safer next step |
|---|---|---|
| Current Gemini 3.5 generation | Gemini 3.5 Flash | Publicly launched and available for real testing |
| Current Pro-family evaluation | Gemini 3.1 Pro | Existing EvoLink Pro route for reasoning and multimodal workloads |
| One integration for multiple routes | Create an EvoLink API key | Keep provider and model selection outside application business logic |
EvoLink is tracking Gemini 3.5 Pro. Support should be described only after upstream access, the request model ID, pricing, route behavior, and at least one real call have been verified.
What developers should verify before adoption
Model ID and channel
Confirm the exact model string and whether it is exposed through the Gemini API, Vertex AI, or another channel. Keep the value in configuration instead of hard-coding a speculative ID.
Pricing and service tier
Wait for an official pricing row. Check standard, batch, caching, priority, and long-context rules only if Google documents them for the actual route.
Capabilities and limits
Verify input modalities, context and output limits, function calling, structured output, grounding, streaming, and file support from the route's capability table.
Production constraints
Check quotas, regions, preview restrictions, latency, error behavior, and deprecation terms. A model appearing in a catalog is the beginning of production evaluation, not the end.
EvoLink route behavior
Before sending production traffic, confirm that the model page, API reference, request model ID, live pricing, and successful-call behavior agree. Do not copy a vendor model name directly into an EvoLink request until EvoLink documents the route.
Prepare without depending on an unreleased model
Teams can make useful progress before launch:
- Keep model selection and fallback policy configurable.
- Save representative coding, agent, document, and multimodal evaluation tasks.
- Record latency, retries, tool errors, token usage, and final task success for current baselines.
- Define the minimum improvement required before moving traffic from Gemini 3.5 Flash or Gemini 3.1 Pro.
- Use a limited canary after launch before changing the default route.
This turns launch day into a controlled route evaluation instead of a rushed integration rewrite.
How this page will change after release
This URL will remain the owner of Gemini 3.5 Pro release-date, API-status, and availability-watch intent. If EvoLink adds a verified Gemini 3.5 Pro route, the dedicated model page will own exact API access, pricing, and model-ID details. This article will add a dated live-status note and link readers to the product page without becoming a second product page.
Related articles and available routes
- Gemini 3.5 Pro vs Gemini 3.5 Flash: Available Now vs Still Pending
- Gemini 3.5 Flash API on EvoLink
- Gemini 3.1 Pro on EvoLink
- Compare the Gemini model family
Sources
- Google: Gemini 3.5 announcement
- Google AI for Developers: Gemini API model catalog
- Google AI for Developers: Gemini API pricing
- Google AI for Developers: Gemini API release notes
- Reuters: reported Gemini 3.5 Pro delay on July 16, 2026
FAQ
Is Gemini 3.5 Pro publicly available in the Gemini API?
No. It is not listed in Google's public Gemini API model catalog as of July 19, 2026. Google has confirmed that the model is in development and used internally, but that is not the same as public API availability.
Did Google officially announce Gemini 3.5 Pro?
Google officially said on May 19 that it was working on Gemini 3.5 Pro, using it internally, and planned a broader rollout. Google has not yet published a public Gemini API route or complete API specification for it.
Has Gemini 3.5 Pro been delayed?
Reuters reported on July 16, citing Bloomberg News, that the launch had been delayed because of internal performance goals, particularly around coding. Treat that explanation as attributed reporting until Google publishes its own detailed status update.
What is the Gemini 3.5 Pro model ID?
gemini-3.5-pro or any rumored variant into production requests.Is Gemini 3.5 Pro available on EvoLink?
Not yet. EvoLink should list it only after upstream access, the request model ID, pricing, route behavior, and a real successful call have been verified.
Which Gemini models can EvoLink users test now?
What should teams prepare before the release?
Keep routing configurable, define representative evaluation tasks, capture current baseline results, and prepare a canary rollout with a fallback. This lets teams evaluate the new route without making an unreleased model a production dependency.
