Gemma 4 26B A4B (Reasoning) logo
Google
•

Open weights model

•

Released April 2026

Gemma 4 26B A4B (Reasoning) API 服务商基准测试与分析

模型比较

分析 Gemma 4 26B A4B (Reasoning) 各 API 服务商的性能指标，包括延迟（首 Token 延迟）、输出速度（每秒输出 token 数）、价格等。参与基准测试的 API 服务商包括 Makora、Novita、Google AI Studio、Cloudflare、DeepInfra、Parasail、Clarifai和GMI (FP8)。

速度最快

#1
MakoraMakora
164.4 t/s
#2
CloudflareCloudflare
108.2 t/s
#3
Google AI StudioGoogle AI Studio
47.0 t/s
#4
ParasailParasail
35.3 t/s
#5
NovitaNovita
34.6 t/s

输出速度

共 8 家服务商

延迟最低

#1
MakoraMakora
13.11 s
#2
CloudflareCloudflare
19.45 s
#3
Google AI StudioGoogle AI Studio
43.62 s
#4
ParasailParasail
58.10 s
#5
NovitaNovita
60.19 s

首个答案 Token 延迟

共 8 家服务商

价格最低

#1
Google AI StudioGoogle AI Studio
$0.00
#2
MakoraMakora
$0.09
#3
DeepInfraDeepInfra
$0.10
#4
ParasailParasail
$0.10
#5
CloudflareCloudflare
$0.12

每 100 万 token 的混合价格

共 8 家服务商

可通过 8 家 API 服务商使用 Gemma 4 26B A4B，每家服务商的性能特征和价格各不相同。下方对比了各服务商的关键指标。

  • 输出速度最快的服务商是 Makora（164.4 t/s）、Cloudflare（108.2 t/s）和Google AI Studio（47.0 t/s）。 各服务商之间的速度差异显著，最快与最慢相差 438%。
  • 延迟方面，Makora（13.11 秒）、Cloudflare（19.45 秒）和Google AI Studio（43.62 秒） 的首个答案 Token 延迟最低。
  • 价格方面，Google AI Studio（0.00）、Makora（0.09）和DeepInfra（0.10） 每 100 万 token 的混合价格最低。
  • Makora 兼具最高速度和最低延迟，性能表现最佳。若要优化成本，Google AI Studio 的价格最具竞争力。

亮点

速度

Output tokens per second · Higher is better

端到端响应时间

Seconds · Lower is better

价格

USD per 1M tokens (blended) · Lower is better

更新：默认性能基准测试工作负载已调整为 1 万个输入 token，以更贴近生产使用场景。你仍可在上方选择其他工作负载。

价格

Pricing: Cache Hit, Input, and Output

Price (USD per M Tokens) · Lower is better · 10,000 input tokens

Price per token for cached prompts (previously processed), typically offering a significant discount compared to regular input price, represented as USD per million tokens. The values shown here are the cache hit price; cache write and cache storage are billed separately and vary by provider — see "Cache pricing by provider" for detail.

Pricing: Blended Price

Blended at 7:2:1 (cache-input-output) · USD per 1M tokens (blended) · Lower is better

Price per token, shown in USD per million tokens. Price is a blend of cache hit, input, and output token prices using the selected ratio (default 7:2:1 cache-input-output).

Pricing: Cache Discount

1 - (cache hit price / input price) · Higher is better

Reduction in input token cost due to cache hit relative to input price. Formula: 1 - (Cache Hit Price per Token / Input Token Price), where cache hit price is the first-party cache hit price or the median provider cache hit price. Note that this discount figure does not account for all costs associated with cache hits, such as cache write and storage costs.

Output Speed vs. Price

Blended at 7:2:1 (cache-input-output) · Output speed: output tokens per second · USD per 1M tokens (blended)
Most attractive quadrant

Tokens per second received while the model is generating tokens (ie. after first chunk has been received from the API for models which support streaming).

速度

按输出速度（每秒 token 数）衡量

Output Speed: Gemma 4 26B A4B (Reasoning)

Output speed: output tokens per second · 10,000 input tokens

Tokens per second received while the model is generating tokens (ie. after first chunk has been received from the API for models which support streaming).

Figures represent performance of the model's first-party API (e.g. OpenAI for o1) or the median across providers where a first-party API is not available (e.g. Meta's Llama models).

Latency vs. Output Speed

Latency: seconds to first token received · Output speed: output tokens per second · 10,000 input tokens
Most attractive quadrant
Pareto line

Tokens per second received while the model is generating tokens (ie. after first chunk has been received from the API for models which support streaming).

延迟

按首 Token 延迟（秒）衡量

首个答案 Token 延迟

Seconds to first token received · Lower is better · 10,000 input tokens

Time to first answer token received, in seconds, after API request sent. For reasoning models, this includes the 'thinking' time of the model before providing an answer. For models which do not support streaming, this represents time to receive the completion.

Figures represent median (P50) measurement over the past 72 hours to reflect sustained changes in performance.

端到端响应时间

Seconds to output 500 tokens, calculated based on time to first token, 'thinking' time for reasoning models, and output speed

端到端响应时间

Seconds to output 500 tokens, including reasoning model 'thinking' time · Lower is better · 10,000 input tokens

Seconds to receive a 500 token response. Key components:

  • Input time: Time to receive the first response token
  • Thinking time (only for reasoning models): Time reasoning models spend outputting tokens to reason prior to providing an answer. Amount of tokens based on the average reasoning tokens across a diverse set of 60 prompts (methodology details).
  • Answer time: Time to generate 500 output tokens, based on output speed

关键比较指标与 API 功能

Makora 标志Makora
1M
开放
$0.02
164
0.95
16.15
12.16
Novita 标志Novita
262k
开放
$0.04
35
2.37
74.64
57.82
Google 标志Google
262k
开放
$0.00
47
1.07
54.26
42.55
Cloudflare 标志Cloudflare
256k
开放
$0.03
108
0.97
24.07
18.48
DeepInfra 标志DeepInfra
262k
开放
$0.03
31
0.76
82.60
65.47
Parasail 标志Parasail
256k
开放
$0.03
35
1.49
72.25
56.61
Clarifai 标志Clarifai
256k
开放
$0.20
--
--
--
--
GMI 标志GMI
1.05M
开放
$0.04
--
--
--
--

常见问题

关于 Gemma 4 26B A4B (Reasoning) 服务商的常见问题

可通过 8 家 API 服务商使用 Gemma 4 26B A4B (Reasoning)：Makora、Novita、Google AI Studio、Cloudflare、DeepInfra、Parasail、Clarifai和GMI (FP8)。每家服务商的性能特征和价格各不相同。

目前可通过我们进行基准测试和跟踪的 8 家 API 服务商使用 Gemma 4 26B A4B (Reasoning)。

按输出速度计算，Gemma 4 26B A4B (Reasoning) 速度最快的服务商是 Makora（164.4 t/s）、Cloudflare（108.2 t/s）和Google AI Studio（47.0 t/s）。输出速度衡量模型开始响应后生成 token 的速度。

Gemma 4 26B A4B (Reasoning) 的首个答案 Token 延迟最低的服务商是 Makora（13.11s）、Cloudflare（19.45s）和Google AI Studio（43.62s）。延迟越低，初始响应越快。

按混合价格计算，Gemma 4 26B A4B (Reasoning) 最实惠的服务商是 Google AI Studio（每 100 万 token $0.00）、Makora（每 100 万 token $0.09）和DeepInfra（每 100 万 token $0.10）。混合价格采用缓存命中/输入/输出 token 为 7:2:1 的比例。

Gemma 4 26B A4B (Reasoning) 输入 token 价格最低的服务商是 Google AI Studio（每 100 万输入 token $0.00）、DeepInfra（每 100 万输入 token $0.07）和Cloudflare（每 100 万输入 token $0.10）。

Gemma 4 26B A4B (Reasoning) 输出 token 价格最低的服务商是 Google AI Studio（每 100 万输出 token $0.00）、Cloudflare（每 100 万输出 token $0.30）和DeepInfra（每 100 万输出 token $0.34）。

Gemma 4 26B A4B (Reasoning) 的输出速度在不同服务商之间差异显著。Makora 最快，为 164.4 t/s，是 DeepInfra（30.5 t/s）的 5.4 倍。

提供 Gemma 4 26B A4B (Reasoning) 的全部 8 家服务商均支持用于结构化输出的 JSON 模式。

提供 Gemma 4 26B A4B (Reasoning) 的全部 8 家服务商均支持函数调用（工具使用）。

对于 Gemma 4 26B A4B (Reasoning)，Makora 兼具最高速度和最低延迟，性能表现最佳。若要优化成本，Google AI Studio 的价格最具竞争力。

选择 Gemma 4 26B A4B (Reasoning) 服务商时，请考虑：输出速度（适合吞吐量密集型任务）、延迟（适合需要快速首次响应的交互式应用）、价格（适合成本敏感型工作负载），以及 JSON 模式或函数调用等 API 功能。

会。服务商性能可能因基础设施变化、负载均衡和更新而随时间波动。我们持续对所有服务商进行基准测试，并在“随时间变化”图表中展示历史性能趋势。

有关 Gemma 4 26B A4B (Reasoning) 的智能、能力、模态及其与其他模型的比较，请参阅模型概览页面。 查看模型概览