DeepSeek
·Released on Dec 1, 2025DeepSeek V3.2 API Benchmarks, Pricing & Provider Data
Compare DeepSeek V3.2 with another model
Choose a model to open its comparison page.
LLM
DeepSeek V3.2 benchmark, API pricing, and provider data cover 451 API providers, with prices starting at $0.0017/request. DeepSeek V3.2 free API options are available from 6 providers. The page also shows measured API speed and first-token latency.
DeepSeek V3.2 is an upgraded V3-series MoE model with stronger reasoning, coding, and math performance, widely available through OpenAI-compatible API relays.
Quality
#28of 100
0.0
LMSpeed score
Speed
46char/s
5.33 s
Cost
#53of 181
$0.0017/ 1M · 8:1 in:out
$0.0003 in · $0.0014 out
Category Performance
Observed-capability estimates with uncertainty reported separately from independent benchmark families.
- Coverage
- 6 / 8
- Methodology
- V3.0
#1Coding51.2RatedGlobal rank #263/4 Measured dimensions
80% interval: 42.7–59.8
code generation59
livecodebench · livecodebench / scicode · scicode
repository engineering39.4
react_native_bench · react_native_evals / vibecode · vibe_code_bench
debugging testing55.3
swe_rebench · swe_rebench
tooling qualityPrior only
#2Reasoning51RatedGlobal rank #373/4 Measured dimensions
80% interval: 42.3–59.7
abstract logicPrior only
scientific causal54.3
critpt · critpt / gpqa · gpqa / hle · hle
multistep constraints57.6
mmlu_pro · mmlu_pro
evidence verification41.1
lcr · lcr
#3Math49Provisional1/4 Measured dimensions
80% interval: 32.9–65.0
foundational mathPrior only
competition mathPrior only
proof frontier49
frontiermath · frontier_math_v2_tier4 / frontiermath · frontier_math_v2_tiers13
applied tool mathPrior only
#4Instruction following46Provisional1/4 Measured dimensions
80% interval: 30.0–62.0
constraint followingPrior only
structured outputPrior only
novel instruction generalization46
ifbench · aa_if_bench
long multiturn instructionPrior only
#5Knowledge39.8Provisional1/4 Measured dimensions
80% interval: 23.7–55.8
broad knowledge39.8
aa_omniscience · aa_omniscience_index
professional knowledgePrior only
factualityPrior only
retrieval open bookPrior only
#6Agents39.1RatedGlobal rank #554/4 Measured dimensions
80% interval: 32.2–46.0
planning39.5
deep_planning · gert_labs
tool use48.8
tau · tau2_bench
environment execution42.3
vita_bench · vita_bench
recovery reliability25.7
claw_eval · claw_eval
No data:MultilingualMultimodal
Specifications
Input and output token limits for this model, plus how it ranks on long-context understanding.
INPUT
163.8Ktokens
≈ 196.6 pages of text
OUTPUT
65.5Ktokens
8K128K1M4M
163.8K
Features
Technical Details
- Input
- Output
- Released
- Dec 2025
- Tokenizer
- DeepSeek
- Architecture
- text->text
- Moderated
- No
- Supported parameters
- frequency_penaltyinclude_reasoninglogit_biaslogprobsmax_tokensmin_ppresence_penaltyreasoningrepetition_penaltyresponse_formatseedstopstructured_outputstemperaturetool_choicetoolstop_ktop_logprobstop_p
Rankings
Excels at
It's decent at
Falls behind in
Detailed scores
Updated: Sep 1, 2026Pricing
undefined metric} other undefined metrics}}
Input price$0.280/M#53 / 181Output price$0.420/M#20 / 181
Pricing
undefined metric} other undefined metrics}}
Benchmark
LMSpeed rank
Score
Source
Updated
Confidence
Agents
V3.0undefined metric} other undefined metrics}} · Rated
Score39.1#5580% interval32.2–46.04/4 Measured dimensionsClaw-Eval40.2#27 / 27VITA-Bench18.5#7 / 9
Agents
V3.0undefined metric} other undefined metrics}} · Rated
Dimensions and evidence
Planning & decomposition39.5
undefined family} other undefined families}} · undefined metric} other undefined metrics}}
deep_planning · gert_labs · z -1.39 · q 1.00
Tool use48.8
undefined family} other undefined families}} · undefined metric} other undefined metrics}}
tau · tau2_bench · z -0.27 · q 1.00
Environment & long-horizon execution42.3
undefined family} other undefined families}} · undefined metric} other undefined metrics}}
vita_bench · vita_bench · z -0.96 · q 1.00
Recovery & completion reliability25.7
undefined family} other undefined families}} · undefined metric} other undefined metrics}}
claw_eval · claw_eval · z -3.00 · q 1.00
Benchmark
LMSpeed rank
Score
Source
Updated
Confidence
Coding
V3.0undefined metric} other undefined metrics}} · Rated
Score51.2#2680% interval42.7–59.83/4 Measured dimensionsLiveCodeBench86.2%#5 / 115SciCode38.9%#98 / 206
Coding
V3.0undefined metric} other undefined metrics}} · Rated
Dimensions and evidence
Code generation59
undefined family} other undefined families}} · undefined metric} other undefined metrics}}
livecodebench · livecodebench · z 1.45 · q 1.00
scicode · scicode · z 0.57 · q 1.00
Repository engineering39.4
undefined family} other undefined families}} · undefined metric} other undefined metrics}}
react_native_bench · react_native_evals · z -0.70 · q 1.00
vibecode · vibe_code_bench · z -1.43 · q 1.00
Debugging & testing55.3
undefined family} other undefined families}} · undefined metric} other undefined metrics}}
swe_rebench · swe_rebench · z 0.78 · q 1.00
Tool-assisted development & qualityPrior only
Benchmark
LMSpeed rank
Score
Source
Updated
Confidence
LiveCodeBenchV3 evidence
LMSpeed rank#5
Score86.2%
Sourceartificialanalysis.ai
UpdatedSep 1, 2026
Confidence4
Reasoning
V3.0undefined metric} other undefined metrics}} · Rated
Score51#3780% interval42.3–59.73/4 Measured dimensionsMMLU-Pro86.2%#18 / 129GPQA84.0%#70 / 213
Reasoning
V3.0undefined metric} other undefined metrics}} · Rated
Dimensions and evidence
Abstract logicPrior only
Scientific & causal reasoning54.3
undefined family} other undefined families}} · undefined metric} other undefined metrics}}
critpt · critpt · z -0.47 · q 1.00
gpqa · gpqa · z 0.67 · q 1.00
hle · hle · z 0.61 · q 1.00
Multi-step constraints57.6
undefined family} other undefined families}} · undefined metric} other undefined metrics}}
mmlu_pro · mmlu_pro · z 0.81 · q 1.00
Evidence integration & verification41.1
undefined family} other undefined families}} · undefined metric} other undefined metrics}}
lcr · lcr · z -1.55 · q 1.00
Benchmark
LMSpeed rank
Score
Source
Updated
Confidence
MMLU-ProV3 evidence
LMSpeed rank#18
Score86.2%
Sourceartificialanalysis.ai
UpdatedSep 1, 2026
Confidence4
Knowledge
V3.0undefined metric} other undefined metrics}} · Provisional
Score39.880% interval23.7–55.81/4 Measured dimensionsArtificial Analysis Intelligence Index25.1#78 / 111AA-GPQA Diamond75.1#81 / 108
Knowledge
V3.0undefined metric} other undefined metrics}} · Provisional
Dimensions and evidence
Broad knowledge39.8
undefined family} other undefined families}} · undefined metric} other undefined metrics}}
aa_omniscience · aa_omniscience_index · z -1.45 · q 1.00
Professional knowledgePrior only
FactualityPrior only
Retrieval & open-book usePrior only
Benchmark
LMSpeed rank
Score
Source
Updated
Confidence
Artificial Analysis Intelligence IndexV3 evidence
LMSpeed rank#78
Score25.1
Sourcebenchlm.ai
UpdatedAug 20, 2026
Confidence1
AA-Omniscience Hallucination Rate
LMSpeed rank#4
Score93.3
Sourcebenchlm.ai
UpdatedAug 20, 2026
Confidence1
Math
V3.0undefined metric} other undefined metrics}} · Provisional
Score4980% interval32.9–65.01/4 Measured dimensionsMath score40.2#42 / 62FrontierMath v2 (Tiers 1-3)22.1#26 / 47
Math
V3.0undefined metric} other undefined metrics}} · Provisional
Dimensions and evidence
Foundational mathPrior only
Competition mathPrior only
Advanced proofs49
undefined family} other undefined families}} · undefined metric} other undefined metrics}}
frontiermath · frontier_math_v2_tier4 · z -0.64 · q 1.00
frontiermath · frontier_math_v2_tiers13 · z -0.06 · q 1.00
Applied & tool-assisted mathPrior only
Benchmark
LMSpeed rank
Score
Source
Updated
Confidence
FrontierMath v2 (Tiers 1-3)V3 evidence
LMSpeed rank#26
Score22.1
Sourcebenchlm.ai
UpdatedAug 20, 2026
Confidence1
FrontierMath v2 (Tier 4)V3 evidence
LMSpeed rank#30
Score2.1
Sourcebenchlm.ai
UpdatedAug 20, 2026
Confidence1
Multilingual
V3.0undefined metric} other undefined metrics}} · No data
No data80% interval30.8–69.20/4 Measured dimensions
Multilingual
V3.0undefined metric} other undefined metrics}} · No data
Dimensions and evidence
Cross-language understandingPrior only
Multilingual generationPrior only
Reasoning transferPrior only
Low-resource robustnessPrior only
Multimodal
V3.0undefined metric} other undefined metrics}} · No data
80% interval30.8–69.20/4 Measured dimensionsDesign Arena Website1186.0#52 / 75
Multimodal
V3.0undefined metric} other undefined metrics}} · No data
Dimensions and evidence
Perception & OCRPrior only
Document & spatial understandingPrior only
Visual reasoningPrior only
Video & grounded actionPrior only
Benchmark
LMSpeed rank
Score
Source
Updated
Confidence
Instruction following
V3.0undefined metric} other undefined metrics}} · Provisional
Score4680% interval30.0–62.01/4 Measured dimensionsAA-IFBench49.0#56 / 84
Instruction following
V3.0undefined metric} other undefined metrics}} · Provisional
Dimensions and evidence
Constraint followingPrior only
Structured outputPrior only
Novel-instruction generalization46
undefined family} other undefined families}} · undefined metric} other undefined metrics}}
ifbench · aa_if_bench · z -0.80 · q 1.00
Multi-turn & long instructionsPrior only
Benchmark
LMSpeed rank
Score
Source
Updated
Confidence
OpenRouter endpoints
15 endpointsThird-party OpenRouter endpoint data, shown separately from LMSpeed measurements. Some 30-minute live performance fields only appear after syncing with an OpenRouter API key.
| Provider endpoint | Input | Output | 1d uptime | 30m latency | 30m throughput | Context / output |
|---|---|---|---|---|---|---|
Baidu baidu/fp8 | $0.280/M | $0.420/M | 100.0% | — | — | undefined tokens / undefined tokens |
Friendli friendli | $0.500/M | $1.50/M | 100.0% | — | — | undefined tokens / undefined tokens |
Novita novita/fp8 | $0.269/M | $0.400/M | 99.9% | — | — | undefined tokens / undefined tokens |
AtlasCloud atlas-cloud/fp8 | $0.260/M | $0.380/M | 99.8% | — | — | undefined tokens / undefined tokens |
DeepInfra deepinfra/fp4 | $0.260/M | $0.380/M | 99.7% | — | — | undefined tokens / undefined tokens |
Google google-vertex | $0.560/M | $1.68/M | 99.5% | — | — | undefined tokens / undefined tokens |
Venice venice | $0.330/M | $0.480/M | 99.4% | — | — | undefined tokens / undefined tokens |
Alibaba alibaba/fp8 | $0.370/M | $1.11/M | 99.3% | — | — | undefined tokens / undefined tokens |
Phala phala | $1/M | $1/M | 99.1% | — | — | undefined tokens / undefined tokens |
StreamLake streamlake/fp8 | $0.214/M | $0.322/M | 99.0% | — | — | undefined tokens / undefined tokens |
SiliconFlow siliconflow/fp8 | $0.259/M | $0.420/M | 98.5% | — | — | undefined tokens / undefined tokens |
DigitalOcean digitalocean | $0.250/M | $0.800/M | 96.2% | — | — | undefined tokens / undefined tokens |
SambaNova sambanova | $3/M | $4.50/M | 96.0% | — | — | undefined tokens / undefined tokens |
GMICloud gmicloud/fp8 | $0.209/M | $0.310/M | 93.0% | — | — | undefined tokens / undefined tokens |
Mara mara | $3/M | $4.50/M | 68.5% | — | — | undefined tokens / undefined tokens |
Pricing Comparison
Compare DeepSeek V3.2 API pricing across 445 providers. Prices range from $0.0017/request to $75.00/M. 乐天图书馆 offers the lowest rate at $0.0017/request. 6 providers offer free API credits or a free tier.
| Provider | Health | Model Variant | Group | Input ($/M) | Output ($/M) | Speed (t/s) | First token | Audit |
|---|---|---|---|---|---|---|---|---|
L1 98% | FAST/deepseek-v3.2 | 翻译 | -27%$0.205/M | -51%$0.205/M | 257.0 t/s | 1.77 s | — | |
L1 98% | deepseek-ai/DeepSeek-V3.2 | 临时渠道 | -99%$0.0037/M Cache read$0.0004/MCache write$0.0037/MCache write 1h$0.0059/M | -99%$0.0058/M | 23.6 t/s | 3.30 s | — | |
L1 98% | Pro/deepseek-ai/DeepSeek-V3.2 | 临时渠道 | -99%$0.0037/M | -99%$0.0058/M | — | — | — | |
L1 98% | deepseek-v3.2 | 临时渠道 | -97%$0.0079/M Cache read$0.0009/M | -95%$0.023/M | — | — | — | |
L1 99% L2 0% | deepseek/deepseek-v3.2 | default | -22%$0.217/M Cache read$0.060/M | -22%$0.326/M | 56.1 t/s | 1.98 s | — | |
L1 100% | deepseek-v3.2 | default | -51%$0.137/M | $1.37/M | 49.1 t/s | 6.90 s | — | |
L1 100% | 次DeepSeek-V3.2 | default | $0.0017/request | - | — | — | — | |
L1 100% | deepseek-v3.2-exp | default | -51%$0.137/M | $0.822/M | — | — | — | |
L1 100% | DeepSeek-V3.2 | default | -51%$0.137/M | $1.37/M | — | — | — | |
L1 100% | DeepSeek-V3.2-Fast | default | $0.479/M | $7.19/M | — | — | — | |
L1 100% | deepseek-v3.2-exp | default | -51%$0.137/M Cache read$0.014/M | -51%$0.205/M | 41.0 t/s | 3.78 s | — | |
L1 100% | deepseek-v3.2 | default | -51%$0.137/M | -51%$0.205/M | 26.4 t/s | 1.22 s | — | |
L1 100% | deepseek-v3.2 | default | -2%$0.274/M | -2%$0.411/M | 36.6 t/s | 1.52 s | — | |
L1 100% | deepseek-v3.2-exp | default | -2%$0.274/M Cache read$0.027/M | -2%$0.411/M | — | — | — | |
L1 100% | deepseek-v3.2 | default | $2.00/M Cache read$0.200/M | $3.00/M | 25.7 t/s | 10.23 s | — | |
L1 100% | deepseek-v3.2 | default | $0.010/request | - | 21.7 t/s | 5.99 s | — | |
L1 100% | deepseek-v3.2-speciale | default | $0.010/request | - | — | — | — | |
L1 100% | deepseek-ai/deepseek-v3.2 | default | $0.010/request | - | — | — | — | |
L1 100% | qwen3.5-omni-plus-thinking-nx | default | $0.010/request | - | — | — | — | |
L1 100% | deepseek-v3.2-exp | default | -51%$0.137/M Cache read$0.014/M | -51%$0.205/M | — | — | — |
Showing 20 model IDs of 143.
Alternatives & Similar Models
Kimi K2.5
kimi-k2-5
Moonshot Kimi K2.5 is an open-weight multimodal agent model with native vision and text input, strong coding performance, and a 256K context window.
197 shared providers
DeepSeek V4 Flash
deepseek-v4-flash
DeepSeek V4 Flash is a fast, cost-efficient language model in the DeepSeek V4 family, optimized for low-latency chat, coding assistance, and high-throughput API workloads while retaining strong reasoning quality.
195 shared providers
DeepSeek V4 Pro
deepseek-v4-pro
DeepSeek V4 Pro is the professional-tier DeepSeek V4 model, targeting frontier reasoning, coding, and agent workflows with maximum capability.
192 shared providers
GLM-5.1
glm-5-1
Zhipu GLM-5.1 is a next-generation GLM model aimed at frontier reasoning, coding, and bilingual agent applications.
190 shared providers
MiniMax M2.5
minimax-m2-5
MiniMax M2.5 is MiniMax's flagship text model for coding and agents, with SOTA-level programming and agentic performance, improved token efficiency, and fast high-TPS API deployment.
190 shared providers
GPT-5.4
gpt-5-4
OpenAI GPT-5.4 extends the GPT-5 family with stronger instruction following, deeper tool use, and improved performance on coding, math, and long-document analysis.
185 shared providers
Frequently Asked Questions
- What benchmark data does DeepSeek V3.2 include?
- LMSpeed shows DeepSeek V3.2 benchmark context, API price, output speed, first-token latency, and provider data across 451 providers when those signals are available.
- What is the DeepSeek V3.2 API price?
- DeepSeek V3.2 has pricing from 451 providers, ranging from $0.0017/request to $75.00/M. 乐天图书馆 has the lowest listed price.
- What does the DeepSeek V3.2 API pricing table include?
- The DeepSeek V3.2 API pricing table compares 451 providers by input price, output price, free tier status, speed, first-token latency, and recent health data when available.
- Which provider has the cheapest DeepSeek V3.2 API pricing?
- 乐天图书馆 currently has the lowest listed DeepSeek V3.2 price at $0.0017/request across 451 providers.
- Can I compare DeepSeek V3.2 API price and speed together?
- Yes. LMSpeed shows DeepSeek V3.2 API price, output speed, first-token latency, and provider health on the same page so you can compare cost and performance together.
- Is DeepSeek V3.2 API free?
- Yes, DeepSeek V3.2 free API options are available through 6 providerundefined other undefined} on LMSpeed, including 兔子API, Zero API, 兔子API, 兔子API, 兔子API. These providers offer free API credits or a free tier with no per-token charges.
- Where can I get DeepSeek V3.2 free API access?
- LMSpeed currently lists 6 free API providerundefined other undefined} for DeepSeek V3.2: 兔子API, Zero API, 兔子API, 兔子API, 兔子API. Check each provider row before using it because free tier limits can change.