# \#thinking

**URL:** https://discuss.ai.google.dev/tag/thinking/411.md

[Latest](https://discuss.ai.google.dev/latest.md) · [Categories](https://discuss.ai.google.dev/categories.md) · [Tags](https://discuss.ai.google.dev/tags.md)

---

## [Gemini 4 Argon: Thinking summaries + token counts missing when using structured JSON](https://discuss.ai.google.dev/t/gemini-4-argon-thinking-summaries-token-counts-missing-when-using-structured-json/191660)

<div class="topic-metadata">

**Author:** [@atlas-founder](https://discuss.ai.google.dev/u/atlas-founder)\
**Replies:** 1\
**Last updated:** [October 23, 2026, 7:38pm UTC](https://discuss.ai.google.dev/t/gemini-4-argon-thinking-summaries-token-counts-missing-when-using-structured-json/191660 "2026-10-23T19:38:01Z")

</div>

(Heads up: I used Gemini to help troubleshoot and write this report, but I’ve personally tested and verified the results below. Thank you!) gemini-4-argon returns thinking summaries and thinking-token counts for normal …

---

## [Gemini 3.8 Flash + File Search: 7–8s TTFT with low thinking (timing breakdown)](https://discuss.ai.google.dev/t/gemini-3-8-flash-file-search-7-8s-ttft-with-low-thinking-timing-breakdown/185764)

<div class="topic-metadata">

**Author:** [@codebird](https://discuss.ai.google.dev/u/codebird)\
**Replies:** 0\
**Last updated:** [September 28, 2026, 11:05am UTC](https://discuss.ai.google.dev/t/gemini-3-8-flash-file-search-7-8s-ttft-with-low-thinking-timing-breakdown/185764 "2026-09-28T11:05:05Z")

</div>

Hi Google team, I’m building a customer-support bot with Gemini 3.8 Flash and File Search, and typically see 7–8 seconds before the first answer text arrives—not time to complete the answer. Could someone from the File …

---

## [thinkingLevel: 'low' producing unpredictable 60k+ thought-token spikes on trivial prompts - systemic billing impact](https://discuss.ai.google.dev/t/thinkinglevel-low-producing-unpredictable-60k-thought-token-spikes-on-trivial-prompts-systemic-billing-impact/169408)

<div class="topic-metadata">

**Author:** [@jmocanu](https://discuss.ai.google.dev/u/jmocanu)\
**Replies:** 1\
**Last updated:** [September 2, 2026, 6:23pm UTC](https://discuss.ai.google.dev/t/thinkinglevel-low-producing-unpredictable-60k-thought-token-spikes-on-trivial-prompts-systemic-billing-impact/169408 "2026-09-02T18:23:10Z")

</div>

I’m filing this as a billing and API correctness issue. I have two days of logged \`usageMetadata\` from a batch job running \`gemini-flash-latest\` (Gemini 3.5 Flash) with \`thinkingConfig: { thinkingLevel: ‘low’ }\` explic…

---

## [Gemini 3.1 Pro in Google AI Studio enters recursive Thinking loops in observer-reference and metafictional interpretation tasks](https://discuss.ai.google.dev/t/gemini-3-1-pro-in-google-ai-studio-enters-recursive-thinking-loops-in-observer-reference-and-metafictional-interpretation-tasks/172478)

<div class="topic-metadata">

**Author:** [@Sixpoint\_Handstand](https://discuss.ai.google.dev/u/Sixpoint_Handstand)\
**Replies:** 0\
**Last updated:** [June 24, 2026, 8:48am UTC](https://discuss.ai.google.dev/t/gemini-3-1-pro-in-google-ai-studio-enters-recursive-thinking-loops-in-observer-reference-and-metafictional-interpretation-tasks/172478 "2026-06-24T08:48:48Z")

</div>

Hi everyone, I encountered this in Google AI Studio using the free tier. Web search / grounding was disabled in one case. The same Google account has Google One and Google Ads subscriptions, but this issue occurred insi…

---

## [Agent replies/thinking/reasoning in IDE missing](https://discuss.ai.google.dev/t/agent-replies-thinking-reasoning-in-ide-missing/170622)

<div class="topic-metadata">

**Author:** [@QueueQ](https://discuss.ai.google.dev/u/QueueQ)\
**Replies:** 5\
**Last updated:** [June 16, 2026, 7:08pm UTC](https://discuss.ai.google.dev/t/agent-replies-thinking-reasoning-in-ide-missing/170622 "2026-06-16T19:08:47Z")

</div>

I’ve noticed that while the agent is working, it will often say some stuff I definitely want to read. Breakthroughs, reasonings, responses, etc., that will only be available during the current Working… task, then it will…

---

## [The thoughts between steps are deleted and disappear when the next step has started:](https://discuss.ai.google.dev/t/the-thoughts-between-steps-are-deleted-and-disappear-when-the-next-step-has-started/167796)

<div class="topic-metadata">

**Author:** [@bjoern](https://discuss.ai.google.dev/u/bjoern)\
**Replies:** 1\
**Last updated:** [May 29, 2026, 6:42pm UTC](https://discuss.ai.google.dev/t/the-thoughts-between-steps-are-deleted-and-disappear-when-the-next-step-has-started/167796 "2026-05-29T18:42:08Z")

</div>

In the Google Antigravity IDE, Create a task where the AI has to read multiple files. During generation, you can see the thought process. When the AI reads or edits the next file, the UI provides no way to see the previ…

---

## [Gemini 3.5 Flash Low thinking option](https://discuss.ai.google.dev/t/gemini-3-5-flash-low-thinking-option/147452)

<div class="topic-metadata">

**Author:** [@Luiz\_Gabriel\_dos\_San](https://discuss.ai.google.dev/u/Luiz_Gabriel_dos_San)\
**Replies:** 0\
**Last updated:** [May 23, 2026, 4:31am UTC](https://discuss.ai.google.dev/t/gemini-3-5-flash-low-thinking-option/147452 "2026-05-23T04:31:31Z")

</div>

Google AI Studio gives developers granular control over reasoning levels, and we are starting to see basic reasoning toggles roll out in the Gemini App. However, the default behavior for Gemini 3.5 Flash is burning thro…

---

## [Gemini-3.1-pro-preview: thoughtsTokenCount missing with thinking\_level=LOW](https://discuss.ai.google.dev/t/gemini-3-1-pro-preview-thoughtstokencount-missing-with-thinking-level-low/135965)

<div class="topic-metadata">

**Author:** [@Montego](https://discuss.ai.google.dev/u/Montego)\
**Replies:** 1\
**Last updated:** [May 15, 2026, 1:02am UTC](https://discuss.ai.google.dev/t/gemini-3-1-pro-preview-thoughtstokencount-missing-with-thinking-level-low/135965 "2026-05-15T01:02:18Z")

</div>

Hi, I’m seeing inconsistent usage metadata for gemini-3.1-pro-preview on Vertex AI. When using: thinking\_level=HIGH → thoughtsTokenCount is present thinking\_level=MEDIUM → thoughtsTokenCount is present thinking\_level…

---

## [Bug: Trajectory permanently poisoned by thinking-terminated assistant turn after stream timeout — every subsequent Claude request fails with HTTP 400](https://discuss.ai.google.dev/t/bug-trajectory-permanently-poisoned-by-thinking-terminated-assistant-turn-after-stream-timeout-every-subsequent-claude-request-fails-with-http-400/144340)

<div class="topic-metadata">

**Author:** [@Davis\_Lin](https://discuss.ai.google.dev/u/Davis_Lin)\
**Replies:** 0\
**Last updated:** [May 11, 2026, 4:21am UTC](https://discuss.ai.google.dev/t/bug-trajectory-permanently-poisoned-by-thinking-terminated-assistant-turn-after-stream-timeout-every-subsequent-claude-request-fails-with-http-400/144340 "2026-05-11T04:21:14Z")

</div>

Tags: bug, api, models, claude, thinking Severity P1 — Blocker. A single interrupted generation permanently breaks the trajectory. The user cannot continue without manually reverting history, and quota is consumed on ev…

---

## [Disable thinking for Gemma 4](https://discuss.ai.google.dev/t/disable-thinking-for-gemma-4/138885)

<div class="topic-metadata">

**Author:** [@letanloc](https://discuss.ai.google.dev/u/letanloc)\
**Replies:** 9\
**Last updated:** [May 10, 2026, 9:29pm UTC](https://discuss.ai.google.dev/t/disable-thinking-for-gemma-4/138885 "2026-05-10T21:29:49Z")

</div>

Hi everyone, I use Gemma 4 via API, I want to disable thinking (because it’s very slow, it spends more than 60s for a simple question) but I cannot find any document to do it. Does anyone know how to disable thinking? …

---

## [Workspace knowledge items exposed in a new workspace](https://discuss.ai.google.dev/t/workspace-knowledge-items-exposed-in-a-new-workspace/122315)

<div class="topic-metadata">

**Author:** [@Sam\_Rexford](https://discuss.ai.google.dev/u/Sam_Rexford)\
**Replies:** 7\
**Last updated:** [May 7, 2026, 11:04pm UTC](https://discuss.ai.google.dev/t/workspace-knowledge-items-exposed-in-a-new-workspace/122315 "2026-05-07T23:04:33Z")

</div>

Hi - recently launched a new project in a new ‘Workspace’ in antigravity. My assumption was this would effectively be a new project with new knowledge items, with data siloed away from other workspaces… In the Antigravit…

---

## [Disappointed with the pro plan. Impractical to use](https://discuss.ai.google.dev/t/disappointed-with-the-pro-plan-impractical-to-use/142501)

<div class="topic-metadata">

**Author:** [@Nelson](https://discuss.ai.google.dev/u/Nelson)\
**Replies:** 2\
**Last updated:** [April 27, 2026, 3:20pm UTC](https://discuss.ai.google.dev/t/disappointed-with-the-pro-plan-impractical-to-use/142501 "2026-04-27T15:20:38Z")

</div>

I want to express my absolute disappointment with antigravity PRO. Seriously, am I literally PAYING to have to wait 6 days (now 5) to use this, every time I use it a little? This is absolutely shameful, Google. I a…

---

## [Antigravity Claude Opus 4.6 Thinking — persistent 503 MODEL\_CAPACITY\_EXHAUSTED (ongoing several hours). Ultra Sub](https://discuss.ai.google.dev/t/antigravity-claude-opus-4-6-thinking-persistent-503-model-capacity-exhausted-ongoing-several-hours-ultra-sub/138497)

<div class="topic-metadata">

**Author:** [@SoSdev](https://discuss.ai.google.dev/u/SoSdev)\
**Replies:** 14\
**Last updated:** [April 21, 2026, 5:08pm UTC](https://discuss.ai.google.dev/t/antigravity-claude-opus-4-6-thinking-persistent-503-model-capacity-exhausted-ongoing-several-hours-ultra-sub/138497 "2026-04-21T17:08:51Z")

</div>

I’ve been hitting repeated 503s on claude-opus-4-6-thinking in Antigravity for several hours now. the errors were happening more frequently over the last few days. tonight, every request fails with the same capacity-exha…

---

## [Can't turn off thinking mode using Gemma4](https://discuss.ai.google.dev/t/cant-turn-off-thinking-mode-using-gemma4/141243)

<div class="topic-metadata">

**Author:** [@Greg\_Obleshchuk](https://discuss.ai.google.dev/u/Greg_Obleshchuk)\
**Replies:** 0\
**Last updated:** [April 20, 2026, 4:13am UTC](https://discuss.ai.google.dev/t/cant-turn-off-thinking-mode-using-gemma4/141243 "2026-04-20T04:13:22Z")

</div>

Hi , I’m calling the gemma4 model both gemma-4-31b-it and gemma-4-26b-a4b-it via rest api and I can’t runoff thinking. this is my url https://generativelanguage.googleapis.com/v1beta/models/gemma-4-31b-it:streamGenerat…

---

## [Big Problem! Gemini 3.0 pro preview thought token exceeding problem](https://discuss.ai.google.dev/t/big-problem-gemini-3-0-pro-preview-thought-token-exceeding-problem/110227)

<div class="topic-metadata">

**Author:** [@komin](https://discuss.ai.google.dev/u/komin)\
**Replies:** 8\
**Last updated:** [March 29, 2026, 11:03am UTC](https://discuss.ai.google.dev/t/big-problem-gemini-3-0-pro-preview-thought-token-exceeding-problem/110227 "2026-03-29T11:03:56Z")

</div>

Hi! Sometimes, when model is set to use a low thinking\_level or thinking\_budget to 128 or 256, it unexpectedly uses around 3,000 thought tokens, even though the task is almost identical to others. This happens with bot…

---

## [Does deleting the "Thinking" block in Google AI Studio affect the next response?](https://discuss.ai.google.dev/t/does-deleting-the-thinking-block-in-google-ai-studio-affect-the-next-response/136274)

<div class="topic-metadata">

**Author:** [@hoshino](https://discuss.ai.google.dev/u/hoshino)\
**Replies:** 1\
**Last updated:** [March 26, 2026, 5:43pm UTC](https://discuss.ai.google.dev/t/does-deleting-the-thinking-block-in-google-ai-studio-affect-the-next-response/136274 "2026-03-26T17:43:45Z")

</div>

Hi everyone, I was reading the official Gemini documentation and came across this statement regarding the stateless nature of the API: “The Gemini API is stateless, so the model treats every API request independently …

---

## [Title: BUG: Pro Tier Quota incorrectly restricted for Claude Sonnet 4.6 Thinking on Antigravity](https://discuss.ai.google.dev/t/title-bug-pro-tier-quota-incorrectly-restricted-for-claude-sonnet-4-6-thinking-on-antigravity/128434)

<div class="topic-metadata">

**Author:** [@Alexandre\_zottino\_ja](https://discuss.ai.google.dev/u/Alexandre_zottino_ja)\
**Replies:** 17\
**Last updated:** [March 13, 2026, 7:41pm UTC](https://discuss.ai.google.dev/t/title-bug-pro-tier-quota-incorrectly-restricted-for-claude-sonnet-4-6-thinking-on-antigravity/128434 "2026-03-13T19:41:51Z")

</div>

Hello, I am a Google AI Pro subscriber, but my Antigravity workspace is incorrectly applying the Free Tier quota limits to my account. I am currently facing an absurd lockout to use the Claude Sonnet 4.5 Thinking model. …

---

## [Follow-up: "Thinking Mode" Instability - Failure Threshold Now at 45k-85k Tokens. Gemini 2.5 Pro](https://discuss.ai.google.dev/t/follow-up-thinking-mode-instability-failure-threshold-now-at-45k-85k-tokens-gemini-2-5-pro/96691)

<div class="topic-metadata">

**Author:** [@Arcadia\_Domus](https://discuss.ai.google.dev/u/Arcadia_Domus)\
**Replies:** 13\
**Last updated:** [February 16, 2026, 9:51pm UTC](https://discuss.ai.google.dev/t/follow-up-thinking-mode-instability-failure-threshold-now-at-45k-85k-tokens-gemini-2-5-pro/96691 "2026-02-16T21:51:51Z")

</div>

Hello everyone, I am the original author of this previous thread - Thinking mode unstable, not working, working 1 time over 20 . I am writing to provide a critical update and a deeper analysis of the “Thinking Mode” ins…

---

## [Preemptive tool calling on gemini 3](https://discuss.ai.google.dev/t/preemptive-tool-calling-on-gemini-3/114514)

<div class="topic-metadata">

**Author:** [@phemmer](https://discuss.ai.google.dev/u/phemmer)\
**Replies:** 0\
**Last updated:** [January 8, 2026, 2:13pm UTC](https://discuss.ai.google.dev/t/preemptive-tool-calling-on-gemini-3/114514 "2026-01-08T14:13:19Z")

</div>

I have an application using the Gemini API with 2.5 models which preemptively injects tool calls. These are tools which actually exist, and I know they contain information the model will use, so rather than wasting time …

---

## [Google Gemini automatically disabling Thinking Module after a few prompts](https://discuss.ai.google.dev/t/google-gemini-automatically-disabling-thinking-module-after-a-few-prompts/88919)

<div class="topic-metadata">

**Author:** [@Lycor](https://discuss.ai.google.dev/u/Lycor)\
**Replies:** 2\
**Last updated:** [November 17, 2025, 3:31pm UTC](https://discuss.ai.google.dev/t/google-gemini-automatically-disabling-thinking-module-after-a-few-prompts/88919 "2025-11-17T15:31:28Z")

</div>

Google Gemini has become horrible. No matter how many times I try to prompt it, it doesn’t enable its thinking mode. It is trying to skimp on tokens. Even after setting a thinking budget, it still does the same thing. …

---

## [Turning off thinking in Gemini Pro 2.5](https://discuss.ai.google.dev/t/turning-off-thinking-in-gemini-pro-2-5/94320)

<div class="topic-metadata">

**Author:** [@RandomGuy](https://discuss.ai.google.dev/u/RandomGuy)\
**Replies:** 3\
**Last updated:** [November 8, 2025, 12:14am UTC](https://discuss.ai.google.dev/t/turning-off-thinking-in-gemini-pro-2-5/94320 "2025-11-08T00:14:19Z")

</div>

Can the Gemini team add an option to turn off the ‘thinking’ process in Gemini pro 2.5? I think for some tasks we need to turn off the reasoning so the model doesn’t drone on when the task doesn’t require thinking.

---

## [How much thinking is enough for agentic applications](https://discuss.ai.google.dev/t/how-much-thinking-is-enough-for-agentic-applications/107282)

<div class="topic-metadata">

**Author:** [@yellowcap](https://discuss.ai.google.dev/u/yellowcap)\
**Replies:** 1\
**Last updated:** [October 15, 2025, 11:25am UTC](https://discuss.ai.google.dev/t/how-much-thinking-is-enough-for-agentic-applications/107282 "2025-10-15T11:25:56Z")

</div>

I am wondering about the optimal thinking budget when building an agent that runs on gemini 2.5 pro. In my case its a simple ReAct agent with only a couple of tools. I wonder if the thinking budget can be optimized both…

---

## [Non-thinking bug, serious issue](https://discuss.ai.google.dev/t/non-thinking-bug-serious-issue/100890)

<div class="topic-metadata">

**Author:** [@RPGBlaster02](https://discuss.ai.google.dev/u/RPGBlaster02)\
**Replies:** 5\
**Last updated:** [October 10, 2025, 3:57am UTC](https://discuss.ai.google.dev/t/non-thinking-bug-serious-issue/100890 "2025-10-10T03:57:38Z")

</div>

How should I explain this… You see, this bug persisted for many, many months regardless of my endless complaints to Logan. For some reason, the Thinking Mode button is grayed out with the Pro 2.5 Model. With a complex Pr…

---

## [Raw chain-of-thought information disclosure vulnerability](https://discuss.ai.google.dev/t/raw-chain-of-thought-information-disclosure-vulnerability/102530)

<div class="topic-metadata">

**Author:** [@Anonymous941](https://discuss.ai.google.dev/u/Anonymous941)\
**Replies:** 0\
**Last updated:** [September 5, 2025, 6:34pm UTC](https://discuss.ai.google.dev/t/raw-chain-of-thought-information-disclosure-vulnerability/102530 "2025-09-05T18:34:57Z")

</div>

As Massive Regression: Detailed Gemini Thinking Process vanished from AI Studio has explained, a few months ago the detailed and very useful thinking process with a terrible summarized version, and didn’t document this i…

---

## [Massive Regression: Detailed Gemini Thinking Process vanished from AI Studio](https://discuss.ai.google.dev/t/massive-regression-detailed-gemini-thinking-process-vanished-from-ai-studio/83916)

<div class="topic-metadata">

**Author:** [@Ji\_LingXuan](https://discuss.ai.google.dev/u/Ji_LingXuan)\
**Replies:** 105\
**Last updated:** [September 5, 2025, 6:35pm UTC](https://discuss.ai.google.dev/t/massive-regression-detailed-gemini-thinking-process-vanished-from-ai-studio/83916 "2025-09-05T18:35:28Z")

</div>

Hi Google AI Team & Community, I’m writing to share my concern about a recent change in Google AI Studio: the detailed thinking process for Gemini models seems to have been removed. Previously, we could see the step-by…

---

## [Thingking Config does not works](https://discuss.ai.google.dev/t/thingking-config-does-not-works/99724)

<div class="topic-metadata">

**Author:** [@aiassist](https://discuss.ai.google.dev/u/aiassist)\
**Replies:** 3\
**Last updated:** [August 26, 2025, 6:36am UTC](https://discuss.ai.google.dev/t/thingking-config-does-not-works/99724 "2025-08-26T06:36:00Z")

</div>

In your official documentation, it says that we can use thinking config to disable thinking for 2.5 flash models Gemini thinking | Gemini API | Google AI for Developers but when i try to use the parameter it says…

---

## [Gemini 2.5 Flash problems while trying to deactivate thinking](https://discuss.ai.google.dev/t/gemini-2-5-flash-problems-while-trying-to-deactivate-thinking/96648)

<div class="topic-metadata">

**Author:** [@Belen\_Cebrian](https://discuss.ai.google.dev/u/Belen_Cebrian)\
**Replies:** 3\
**Last updated:** [August 18, 2025, 7:42pm UTC](https://discuss.ai.google.dev/t/gemini-2-5-flash-problems-while-trying-to-deactivate-thinking/96648 "2025-08-18T19:42:56Z")

</div>

I have a bunch of scripts still using Gemini gemini-1.5-flash-002. I’ve tried to migrate to 2.5 Flash before, but the thinking mode being activated by default results in a much higher cost than I’m able to assume. Probl…

---

## [Thinking mode unstable, not working, working 1 time over 20](https://discuss.ai.google.dev/t/thinking-mode-unstable-not-working-working-1-time-over-20/92285)

<div class="topic-metadata">

**Author:** [@Arcadia\_Domus](https://discuss.ai.google.dev/u/Arcadia_Domus)\
**Replies:** 31\
**Last updated:** [August 6, 2025, 9:38pm UTC](https://discuss.ai.google.dev/t/thinking-mode-unstable-not-working-working-1-time-over-20/92285 "2025-08-06T21:38:40Z")

</div>

I tried everything. I tried putting it into the system prompt. I tried asking it to actively use the thinking mode. I really tried everything, trust me. It’s not working with high-context chats, nor with -100K context ch…

---

## [Whats coming as AI thinks](https://discuss.ai.google.dev/t/whats-coming-as-ai-thinks/94524)

<div class="topic-metadata">

**Author:** [@Ashmika\_Agarwal](https://discuss.ai.google.dev/u/Ashmika_Agarwal)\
**Replies:** 2\
**Last updated:** [August 3, 2025, 6:36am UTC](https://discuss.ai.google.dev/t/whats-coming-as-ai-thinks/94524 "2025-08-03T06:36:42Z")

</div>

Continuing the discussion from Can AI really think?: Yes it can .. i ha e it. Although they may or may not be resonable. I am working on concept of next req for them in coming days.

---

## [Improving Clarity with 'Quote Important Thought'](https://discuss.ai.google.dev/t/improving-clarity-with-quote-important-thought/95933)

<div class="topic-metadata">

**Author:** [@RandomGuy](https://discuss.ai.google.dev/u/RandomGuy)\
**Replies:** 3\
**Last updated:** [July 31, 2025, 6:37am UTC](https://discuss.ai.google.dev/t/improving-clarity-with-quote-important-thought/95933 "2025-07-31T06:37:23Z")

</div>

Can the dev team revert back to “Thinking” or change “Summarize Thought” to “Quote Important Thought” instead? I read “Summarize” and I don’t understand what’s missing when it’s rephrased.

[Next page](https://discuss.ai.google.dev/tag/thinking/411.md?match_all_tags=true&page=1&tags%5B%5D=thinking)
