# Gemini 3 FlashHigh

> Source: Benchize (https://benchize.com) — https://benchize.com/models/gemini-3-flash-high. When citing or reusing this content, credit "Benchize" and link to https://benchize.com/models/gemini-3-flash-high.

Google

Canonical URL: https://benchize.com/models/gemini-3-flash-high
Last updated: Thu, 22 Oct 2026 08:00:11 GMT

Reproduction and quotation must name Benchize, link to the canonical HTML URL, and retain individual source citations.

Google

## Benchmarks

| Benchmark | Score | Rank |  |
| --- | --- | --- | --- |
| [τ-bench telecom](https://benchize.com/benchmarks/sierra%3Atelecom.md) | 91.23% | 3 / 8 |  |
| ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `pass_1` \| 91.23 \| percent \| consistent_success_across_k_trials \| Higher is better \| \| `pass_2` \| 83.48 \| percent \| consistent_success_across_k_trials \| Higher is better \| \| `pass_3` \| 76.54 \| percent \| consistent_success_across_k_trials \| Higher is better \| \| `pass_4` \| 70.18 \| percent \| consistent_success_across_k_trials \| Higher is better \| ### Configuration benchmarkVersion 1.0.1 harness tau-bench methodology {"evaluation_date":"2026-03-02","notes":"Evaluated using vertex_ai/gemini-3-flash-preview with reasoning_effort: high. User simulator: gpt-5.2 with reasoning_effort: low. 4 trials. Seed: 123. Telecom domain had 20 simulations terminated due to too_many_errors (all scored as 0). Banking domain had 52 simulations terminated due to max_steps (all scored as 0). Banking domain evaluated with terminal-based agentic search retrieval (terminal_use). banking_knowledge re-graded under tau2-bench v1.0.1 grading fixes on 2026-07-15 (extra-read logging #329, numeric-argument normalization #397, T077-T086 gold trajectories #402); scores prior to v1.0.1 are not comparable.","tau2_bench_version":"1.0.1","user_simulator":"gpt-5.2","verification":{"details":"Standard tau-bench scaffold.","modified_prompts":false,"omitted_questions":false}} modality text Reasoning effort high retrieval Not reported sourceSubmission gemini-3-flash_sierra_2026-03-02 ### Source [Original source](https://sierra-tau-bench-public.s3.us-west-2.amazonaws.com/submissions/gemini-3-flash_sierra_2026-03-02/submission.json) Source sierra Original model label Gemini 3 Flash Evaluated 2026-03-02 Observed 2026-10-21T16:54:48.424Z Evidence SHA-256 `d1b7d1fc5d789662fb6eeb4a5ec01948295db397d2313df42c43c7fd8caf3637` Source locator `/results/telecom` |  |  |  |
| [τ-bench airline](https://benchize.com/benchmarks/sierra%3Aairline.md) | 82.5% | 3 / 8 |  |
| ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `pass_1` \| 82.5 \| percent \| consistent_success_across_k_trials \| Higher is better \| \| `pass_2` \| 76.33 \| percent \| consistent_success_across_k_trials \| Higher is better \| \| `pass_3` \| 72 \| percent \| consistent_success_across_k_trials \| Higher is better \| \| `pass_4` \| 68 \| percent \| consistent_success_across_k_trials \| Higher is better \| ### Configuration benchmarkVersion 1.0.1 harness tau-bench methodology {"evaluation_date":"2026-03-02","notes":"Evaluated using vertex_ai/gemini-3-flash-preview with reasoning_effort: high. User simulator: gpt-5.2 with reasoning_effort: low. 4 trials. Seed: 123. Telecom domain had 20 simulations terminated due to too_many_errors (all scored as 0). Banking domain had 52 simulations terminated due to max_steps (all scored as 0). Banking domain evaluated with terminal-based agentic search retrieval (terminal_use). banking_knowledge re-graded under tau2-bench v1.0.1 grading fixes on 2026-07-15 (extra-read logging #329, numeric-argument normalization #397, T077-T086 gold trajectories #402); scores prior to v1.0.1 are not comparable.","tau2_bench_version":"1.0.1","user_simulator":"gpt-5.2","verification":{"details":"Standard tau-bench scaffold.","modified_prompts":false,"omitted_questions":false}} modality text Reasoning effort high retrieval Not reported sourceSubmission gemini-3-flash_sierra_2026-03-02 ### Source [Original source](https://sierra-tau-bench-public.s3.us-west-2.amazonaws.com/submissions/gemini-3-flash_sierra_2026-03-02/submission.json) Source sierra Original model label Gemini 3 Flash Evaluated 2026-03-02 Observed 2026-10-21T16:54:48.424Z Evidence SHA-256 `d1b7d1fc5d789662fb6eeb4a5ec01948295db397d2313df42c43c7fd8caf3637` Source locator `/results/airline` |  |  |  |
| [τ-bench banking knowledge](https://benchize.com/benchmarks/sierra%3Abanking_knowledge.md) | 27.32% | 11 / 19 |  |
| ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `pass_1` \| 27.32 \| percent \| consistent_success_across_k_trials \| Higher is better \| \| `pass_2` \| 15.46 \| percent \| consistent_success_across_k_trials \| Higher is better \| \| `pass_3` \| 10.31 \| percent \| consistent_success_across_k_trials \| Higher is better \| \| `pass_4` \| 7.22 \| percent \| consistent_success_across_k_trials \| Higher is better \| ### Configuration benchmarkVersion 1.0.1 harness tau-bench methodology {"evaluation_date":"2026-03-02","notes":"Evaluated using vertex_ai/gemini-3-flash-preview with reasoning_effort: high. User simulator: gpt-5.2 with reasoning_effort: low. 4 trials. Seed: 123. Telecom domain had 20 simulations terminated due to too_many_errors (all scored as 0). Banking domain had 52 simulations terminated due to max_steps (all scored as 0). Banking domain evaluated with terminal-based agentic search retrieval (terminal_use). banking_knowledge re-graded under tau2-bench v1.0.1 grading fixes on 2026-07-15 (extra-read logging #329, numeric-argument normalization #397, T077-T086 gold trajectories #402); scores prior to v1.0.1 are not comparable.","tau2_bench_version":"1.0.1","user_simulator":"gpt-5.2","verification":{"details":"Standard tau-bench scaffold.","modified_prompts":false,"omitted_questions":false}} modality text Reasoning effort high retrieval terminal sourceSubmission gemini-3-flash_sierra_2026-03-02 ### Source [Original source](https://sierra-tau-bench-public.s3.us-west-2.amazonaws.com/submissions/gemini-3-flash_sierra_2026-03-02/submission.json) Source sierra Original model label Gemini 3 Flash Evaluated 2026-03-02 Observed 2026-10-21T16:54:48.424Z Evidence SHA-256 `d1b7d1fc5d789662fb6eeb4a5ec01948295db397d2313df42c43c7fd8caf3637` Source locator `/results/banking_knowledge` |  |  |  |
| [τ-bench retail](https://benchize.com/benchmarks/sierra%3Aretail.md) | 76.75% | 4 / 8 |  |
| ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `pass_1` \| 76.75 \| percent \| consistent_success_across_k_trials \| Higher is better \| \| `pass_2` \| 65.94 \| percent \| consistent_success_across_k_trials \| Higher is better \| \| `pass_3` \| 57.89 \| percent \| consistent_success_across_k_trials \| Higher is better \| \| `pass_4` \| 51.75 \| percent \| consistent_success_across_k_trials \| Higher is better \| ### Configuration benchmarkVersion 1.0.1 harness tau-bench methodology {"evaluation_date":"2026-03-02","notes":"Evaluated using vertex_ai/gemini-3-flash-preview with reasoning_effort: high. User simulator: gpt-5.2 with reasoning_effort: low. 4 trials. Seed: 123. Telecom domain had 20 simulations terminated due to too_many_errors (all scored as 0). Banking domain had 52 simulations terminated due to max_steps (all scored as 0). Banking domain evaluated with terminal-based agentic search retrieval (terminal_use). banking_knowledge re-graded under tau2-bench v1.0.1 grading fixes on 2026-07-15 (extra-read logging #329, numeric-argument normalization #397, T077-T086 gold trajectories #402); scores prior to v1.0.1 are not comparable.","tau2_bench_version":"1.0.1","user_simulator":"gpt-5.2","verification":{"details":"Standard tau-bench scaffold.","modified_prompts":false,"omitted_questions":false}} modality text Reasoning effort high retrieval Not reported sourceSubmission gemini-3-flash_sierra_2026-03-02 ### Source [Original source](https://sierra-tau-bench-public.s3.us-west-2.amazonaws.com/submissions/gemini-3-flash_sierra_2026-03-02/submission.json) Source sierra Original model label Gemini 3 Flash Evaluated 2026-03-02 Observed 2026-10-21T16:54:48.424Z Evidence SHA-256 `d1b7d1fc5d789662fb6eeb4a5ec01948295db397d2313df42c43c7fd8caf3637` Source locator `/results/retail` |  |  |  |

---
Cite as: Gemini 3 FlashHigh — Benchize, https://benchize.com/models/gemini-3-flash-high (retrieved 2026-10-22)