NOUMENOpen app
NOUMEN

Google · Efficient

Gemini 2.5 Flash Preview (09/2025) – Thinking

Added Sep 25, 2025

Compare models

Choose up to 3 models to compare.1 of 3 chosen

Loading models…

Use in the app

Same checkpoint with thinking enabled by default for deeper reasoning and stepwise analysis on complex tasks. Routed to Gemini 2.5 Flash Thinking (stable).

  • Reasoning
  • Tools
  • Vision
  • Structured output
  • PDF input
  • Audio input
Input
Text, Images, Audio
Output
Text
Reasoning effort
low · medium · high
Discovery tier
Efficient

Pricing

USD per 1M tokens
Input
$0.30
Output
$2.50
Cache read
$0.03

Reference prices as published by the model provider. API example

Specifications

Context window
1M
Max output
65.5K

Side by side

Gemini 2.5 Flash Preview (09/2025) – Thinking compared

Specifications, prices, speed and benchmark scores next to each other. Benchmarks below show every compared model too.

Compare models

Choose up to 3 models to compare.1 of 3 chosen

Loading models…

Gemini 2.5 Flash Preview (09/2025) – ThinkingGoogleThis modelKimi K2 0905MoonshotRemove
ProviderGoogleMoonshot
Discovery tierEfficientOpen weights
AddedSep 25, 2025Sep 25, 2025
Limits
Context window1.05MBest262.14K
Max output65.54K100.35KBest
Average output—677 tokens
Pricing, USD per 1M tokens
Input$0.30Best$0.40
Output$2.50$1.80Best
Cache read$0.03—
Performance
Throughput—55.2 tok/s
Time to first token—3.2 s
Request success (24h)99.3%100.0%Best
Benchmark indexes
Intelligence Index15.5Best15.3
Math Index78.3Best57.3
Capabilities
ReasoningYesNo
ToolsYesYes
VisionYesNo
Structured outputYesYes
PDF inputYesNo
Audio inputYesNo
Reasoning effortlow, medium, high—

Public evaluations

Benchmarks

Independent evaluation results, with how each score ranks among the models in the Noumen catalog.

  • Gemini 2.5 Flash Preview (09/2025) – Thinking
  • Kimi K2 0905

Intelligence Index

  • Gemini 2.5 Flash Preview (09/2025) – Thinking15.5
  • Kimi K2 090515.3

Gemini 2.5 Flash Preview (09/2025) – Thinking: Better than 39% of 326 scored models

3.8Catalog models by score55.1

Indexes

  • Math Index

    Composite of the math evaluations

    • Gemini 2.5 Flash Preview (09/2025) – Thinking: 78.3Top
    • Kimi K2 0905: 57.3

Document reasoning

  • AA-LCRv1.1

    Reasoning across long documents

    • Gemini 2.5 Flash Preview (09/2025) – Thinking: 71.0%Top
    • Kimi K2 0905: 53.0%

Reasoning and knowledge

  • Humanity’s Last Exam

    Expert-level questions across academic fields

    • Gemini 2.5 Flash Preview (09/2025) – Thinking: 13.8%Top
    • Kimi K2 0905: 6.4%

  • GPQA Diamond

    Graduate-level science questions

    • Gemini 2.5 Flash Preview (09/2025) – Thinking: 79.3%Top
    • Kimi K2 0905: 76.7%

  • MMLU-Pro

    Multi-subject knowledge and reasoning

    • Gemini 2.5 Flash Preview (09/2025) – Thinking: 84.2%Top
    • Kimi K2 0905: 81.9%

  • IFBench

    Following precise output instructions

    • Gemini 2.5 Flash Preview (09/2025) – Thinking: 52.3%Top
    • Kimi K2 0905: 41.7%

Coding

  • LiveCodeBench

    Recent competitive programming problems

    • Gemini 2.5 Flash Preview (09/2025) – Thinking: 71.3%Top
    • Kimi K2 0905: 61.0%

Math

  • AIME 2025

    Competition mathematics, 2025 problems

    • Gemini 2.5 Flash Preview (09/2025) – Thinking: 78.3%Top
    • Kimi K2 0905: 57.3%

Legacy evaluations

  • Terminal-Bench Hard

    Earlier terminal task suite

    • Gemini 2.5 Flash Preview (09/2025) – Thinking: 16.7%
    • Kimi K2 0905: 23.5%Top

  • τ²-Bench Telecom

    Tool use in a customer-support simulation

    • Gemini 2.5 Flash Preview (09/2025) – Thinking: 45.6%
    • Kimi K2 0905: 73.4%Top

  • AA-LCR (unversioned)

    Long-context reasoning, earlier grading

    • Gemini 2.5 Flash Preview (09/2025) – Thinking: 71.0%Top
    • Kimi K2 0905: 53.0%

Scores from Artificial Analysis, updated Oct 26, 2026. Artificial Analysis

Live routing data

Performance and reliability

Recent averages across requests routed for this model. Your own speed depends on prompt length, reasoning effort and load.

Request success (24h)
99.3%
Reliable
Hourly request success · last 72 hoursLowest hour 99.0%
  • Healthy (98% or more)
  • Degraded (90–98%)
  • Down (under 90%)
  • No requests

Developers

Use it through the API

Noumen’s API is OpenAI-compatible: any Chat Completions client works with a base URL and a key.

Read the quickstart
curl https://www.noumen.xyz/api/v1/chat/completions \
  -H "Authorization: Bearer $NOUMEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gemini-2.5-flash-preview-09-2025-thinking",
    "reasoning_effort": "medium",
    "messages": [{ "role": "user", "content": "Explain this stack trace." }]
  }'
import os
from openai import OpenAI

client = OpenAI(base_url="https://www.noumen.xyz/api/v1", api_key=os.environ["NOUMEN_API_KEY"])
reply = client.chat.completions.create(
    model="gemini-2.5-flash-preview-09-2025-thinking",
    messages=[{"role": "user", "content": "Explain this stack trace."}],
)
print(reply.choices[0].message.content)

Google

Other Google models, starting with the same family.