Databricks:模型智能、性能与价格

Databricks
Databricks

分析 Databricks 各模型的关键指标,包括质量、价格、输出速度、延迟、上下文窗口等。 本分析旨在帮助你根据使用场景,选择 Databricks 提供的最佳模型。

(te)

最智能

#1
Kimi K3 (max)
Kimi K3 (max)
57
#2
GLM-5.2 (max)
GLM-5.2 (max)
51
#3
GPT-5.2 (xhigh)
GPT-5.2 (xhigh)
42
#4
Kimi K2.7 Code
Kimi K2.7 Code
42
#5
Claude Sonnet 5 (Non-reasoning)
Claude Sonnet 5 (Non-reasoning)
42

Intelligence Index

共 19 个模型

速度最快

#1
GLM-5.2 (max)
GLM-5.2 (max)
357 t/s
#2
gpt-oss-120b (high)
gpt-oss-120b (high)
328 t/s
#3
gpt-oss-120b (low)
gpt-oss-120b (low)
327 t/s
#4
gpt-oss-20b (high)
gpt-oss-20b (high)
311 t/s
#5
gpt-oss-20b (low)
gpt-oss-20b (low)
291 t/s

输出速度

共 19 个模型

价格最低

#1
GPT-5 nano (high)
GPT-5 nano (high)
$0.09
#2
gpt-oss-20b (high)
gpt-oss-20b (high)
$0.09
#3
gpt-oss-20b (low)
gpt-oss-20b (low)
$0.09
#4
Llama 3.1 8B
Llama 3.1 8B
$0.18
#5
Gemma 3 12B
Gemma 3 12B
$0.18

每 100 万 token 的混合价格

共 19 个模型

表示推理模型

Databricks 提供 19 个模型,每个模型的智能、性能和价格特征各不相同。 下方对比了各模型的关键指标。

  • 智能方面,Databricks 上表现最好的模型是 Kimi K3 (max)(57)、GLM-5.2 (max)(51)和GPT-5.2 (xhigh)(42)。
  • 输出速度方面,最快的模型是 GLM-5.2 (max)(357 t/s)、gpt-oss-120b (high)(328 t/s)和gpt-oss-120b (low)(327 t/s)。
  • 延迟方面,Llama 4 Maverick(0.72 秒)、Gemma 3 12B(0.85 秒)和Llama 3.1 8B(0.90 秒) 的首个答案 Token 延迟最低。
  • 价格方面,GPT-5 nano (high)($0.09)、gpt-oss-20b (high)($0.09)和gpt-oss-20b (low)($0.09) 每 100 万 token 的混合价格最低。 各模型价格最多相差 2.2 倍。
  • 上下文窗口方面,GLM-5.2 (max)(1M)、Claude 4.5 Sonnet(1M)和Claude Sonnet 5 (Non-reasoning)(524k) 支持 Databricks 上最大的上下文窗口。

亮点

Artificial Analysis Intelligence Index · Higher is better
Output tokens per second · Higher is better
USD per 1M tokens (blended) · Lower is better

智能评测

Artificial Analysis Intelligence Index

Artificial Analysis Intelligence Index v4.1 incorporates 9 evaluations: GDPval-AA v2, 𝜏³-Banking, Terminal-Bench v2.1, SciCode, Humanity's Last Exam, GPQA Diamond, CritPt, AA-Omniscience, AA-LCR
Estimate (independent evaluation forthcoming)
Reasoning models are indicated by a lightbulb icon

Artificial Analysis Intelligence Index v4.1 includes: GDPval-AA v2, 𝜏³-Banking, Terminal-Bench v2.1, SciCode, Humanity's Last Exam, GPQA Diamond, CritPt, AA-Omniscience, AA-LCR. See Intelligence Index methodology for further details, including a breakdown of each evaluation and how we run them.

Intelligence Evaluations

Intelligence evaluations measured independently by Artificial Analysis · Higher is better

Agentic real-world work tasks, (Elo-500)/2000

Agentic tool use

Agentic coding & terminal use

Coding

Reasoning & knowledge

Scientific reasoning

Physics reasoning

Long context reasoning

Agentic knowledge work, Elo

Agentic SaaS workflows

Legal agentic work, criterion pass rate

Agentic business operations

Instruction following

Long-horizon agentic tasks

Kubernetes incident root-cause analysis

Visual reasoning

Reasoning models are indicated by a lightbulb icon

While model intelligence generally translates across use cases, specific evaluations may be more relevant for certain use cases.

Artificial Analysis Intelligence Index v4.1 includes: GDPval-AA v2, 𝜏³-Banking, Terminal-Bench v2.1, SciCode, Humanity's Last Exam, GPQA Diamond, CritPt, AA-Omniscience, AA-LCR. See Intelligence Index methodology for further details, including a breakdown of each evaluation and how we run them.

Intelligence Index vs. Price

Blended at 7:2:1 (cache-input-output) · USD per 1M tokens (blended)
Most attractive quadrant
Pareto line
Reasoning models are indicated by a lightbulb icon

While higher intelligence models are typically more expensive, they do not all follow the same price-quality curve.

上下文窗口

Context Window

Context window: tokens limit · Higher is better
Reasoning models are indicated by a lightbulb icon

Larger context windows are relevant to RAG (Retrieval Augmented Generation) LLM workflows which typically involve reasoning and information retrieval of large amounts of data.

Maximum number of combined input & output tokens. Output tokens commonly have a significantly lower limit (varied by model).

价格

Intelligence Index vs. Price

Blended at 7:2:1 (cache-input-output) · USD per 1M tokens (blended)
Most attractive quadrant
Pareto line
Reasoning models are indicated by a lightbulb icon

While higher intelligence models are typically more expensive, they do not all follow the same price-quality curve.

性能摘要

Output Speed vs. Price

Output speed: output tokens per second · USD per 1M tokens (blended)
Most attractive quadrant
Pareto line
Reasoning models are indicated by a lightbulb icon

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).

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).

速度

按输出速度(每秒 token 数)衡量

Output Speed

Output tokens per second · Higher is better
Reasoning models are indicated by a lightbulb icon

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).

延迟

按首 Token 延迟(秒)衡量

Latency: Time To First Answer Token

Seconds to first answer token received · Accounts for reasoning model 'thinking' time
Reasoning models are indicated by a lightbulb icon

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.

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

End-to-End Response Time vs. Price

End-to-end response time: end-to-end seconds to output 500 tokens · USD per 1M tokens (blended)
Most attractive quadrant
Pareto line
Reasoning models are indicated by a lightbulb icon

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).

进一步分析
Kimi 标志
Kimi K3 (max)
205k
开放
57
$0.81
186
1.08
14.55
10.77
Z AI 标志
GLM-5.2 (max)
1M
开放
51
--
357
1.00
8.00
5.60
OpenAI 标志
GPT-5.2 (xhigh)
400k
专有
42*
--
78
184.85
191.26
--
Kimi 标志
Kimi K2.7 Code
262k
开放
42
$0.20
108
0.96
26.12
20.55
Anthropic 标志
Claude Sonnet 5 (Non-reasoning)
524k
专有
42
$0.48
64
2.10
9.97
--
Google 标志
Gemini 3 Pro Preview (high)
1M
专有
40*
--
--
--
--
--
OpenAI 标志
GPT-5.1 (high)
272k
专有
37
--
104
34.14
38.95
--
OpenAI 标志
GPT-5 (high)
400k
专有
35
--
92
115.99
121.45
--
Anthropic 标志
Claude Opus 4.5
200k
专有
35*
--
47
2.19
12.88
--
Anthropic 标志
Claude 4.5 Sonnet
1M
专有
29*
--
50
2.26
12.19
--
Anthropic 标志
Claude 4.1 Opus
200k
专有
28*
--
--
--
--
--
Anthropic 标志
Claude 4 Sonnet
1M
专有
25*
--
--
--
--
--
OpenAI 标志
GPT-5 mini (high)
400k
专有
25
$0.07
90
114.17
119.76
--
OpenAI 标志
gpt-oss-120b (high)
128k
开放
24
$0.08
328
0.64
8.28
6.11
OpenAI 标志
GPT-5 nano (high)
400k
专有
20*
--
142
113.89
117.42
--
OpenAI 标志
gpt-oss-120b (low)
128k
开放
15
$0.02
327
0.64
8.27
6.11
OpenAI 标志
gpt-oss-20b (high)
131k
开放
15
$0.03
311
0.94
8.97
6.42
OpenAI 标志
gpt-oss-20b (low)
131k
开放
14*
--
291
0.85
9.45
6.88
Meta 标志
Llama 4 Maverick
131k
开放
14
$0.06
85
0.72
6.58
--
Meta 标志
Llama 3.3 70B
128k
开放
9
$0.08
67
1.22
8.73
--
Meta 标志
Llama 3.1 405B
128k
开放
9*
--
--
--
--
--
Meta 标志
Llama 3.1 8B
128k
开放
8
--
197
0.90
3.44
--
Google 标志
Gemma 3 12B
128k
开放
6
$0.26
91
0.85
6.34
--

关键定义

Maximum number of combined input & output tokens. Output tokens commonly have a significantly lower limit (varied by model).

常见问题

关于 Databricks 的常见问题

Databricks 上最智能的模型是 Kimi K3 (max),Intelligence Index 得分为 57。

按输出速度计算,Databricks 上最快的模型是 GLM-5.2 (max),速度为每秒 357.3 个 token。

Databricks 上首个答案 Token 延迟最低的模型是 Llama 4 Maverick,延迟为 0.72 秒。延迟越低,初始响应越快。

按混合价格计算,Databricks 上最实惠的模型是 GPT-5 nano (high),每 100 万 token 的价格为 $0.09(缓存命中/输入/输出比例为 7:2:1)。

Databricks 上各模型价格最多相差 82 倍,从最实惠的 GPT-5 nano (high)(每 100 万 token $0.09)到最昂贵的 Claude Opus 4.5(每 100 万 token $7.00)。

是,Databricks 提供兼容 OpenAI 的 API,可以轻松从 OpenAI 切换,或继续使用现有的 OpenAI SDK 集成。

Databricks 上 19 个模型中,有 16 个支持用于结构化输出的 JSON 模式。

是,Databricks 上全部 19 个模型均支持函数调用(工具使用)。

是,Databricks 提供 12 个推理模型:Kimi K3 (max)GLM-5.2 (max)GPT-5.2 (xhigh)Kimi K2.7 CodeGPT-5.1 (high)GPT-5 (high)GPT-5 mini (high)gpt-oss-120b (high)GPT-5 nano (high)gpt-oss-120b (low)gpt-oss-20b (high)gpt-oss-20b (low)。推理模型会先进行扩展思考来解决复杂问题,再给出答案。

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。

选择 Databricks 上的模型时,请考虑:智能(适合质量敏感型任务)、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),以及上下文窗口大小、JSON 模式或函数调用支持等功能。