Qwen
·Released on Aug 14, 2026Qwen3.8 27B API Benchmarks, Pricing & Provider Data
Compare Qwen3.8 27B with another model
Choose a model to open its comparison page.
LLM
Qwen3.8 27B is an open-weight dense vision-language model from Qwen. It is suited for coding, professional workflows, research, multimodal interaction, and long-running agent tasks, with flexible thin...
Quality
#24of 100
68.0
LMSpeed score
Category Performance
Observed-capability estimates with uncertainty reported separately from independent benchmark families.
- Coverage
- 6 / 8
- Methodology
- V3.0
#1Reasoning58.1Provisional1/4 Measured dimensions
80% interval: 44.2–72.0
abstract logicPrior only
scientific causal58.1
gpqa · gpqa / hle · hle / hle · hle_no_tools
multistep constraintsPrior only
evidence verificationPrior only
#2Multimodal57.7Estimated2/4 Measured dimensions
80% interval: 46.4–69.0
perception ocrPrior only
document spatial59.4
charxiv · charxiv
visual reasoning56
erqa · erqa / mathvision · math_vision
video actionPrior only
#3Agents57.2Estimated2/4 Measured dimensions
80% interval: 45.3–69.1
planningPrior only
tool usePrior only
environment execution62
osworld · os_world_verified
recovery reliability52.4
jobbench · job_bench
#4Coding56.2Estimated2/4 Measured dimensions
80% interval: 45.6–66.9
code generation59.9
livecodebench · live_code_bench_v6 / scicode · scicode
repository engineering52.6
nl2repo · nl2_repo / swe_pro · swe_pro
debugging testingPrior only
tooling qualityPrior only
#5Instruction following56.1Provisional1/4 Measured dimensions
80% interval: 40.1–72.2
constraint followingPrior only
structured outputPrior only
novel instruction generalization56.1
ifbench · if_bench
long multiturn instructionPrior only
#6Knowledge41.5Provisional1/4 Measured dimensions
80% interval: 25.4–57.5
broad knowledge41.5
benchlm_category_knowledge · benchlm_category_knowledge
professional knowledgePrior only
factualityPrior only
retrieval open bookPrior only
No data:MathMultilingual
Specifications
Input and output token limits for this model, plus how it ranks on long-context understanding.
INPUT
1Mtokens
≈ 1.2K pages of text
OUTPUT
131.1Ktokens
8K128K1M4M
1M
Features
Technical Details
- Input
- Output
- Total parameters
- 27B
- Released
- Aug 2026
- Tokenizer
- Qwen
- Architecture
- text+image+video->text
- Moderated
- No
- Supported parameters
- frequency_penaltyinclude_reasoninglogit_biaslogprobsmax_tokenspresence_penaltyreasoningreasoning_effortrepetition_penaltyresponse_formatseedstopstructured_outputstemperaturetool_choicetoolstop_ktop_logprobstop_p
Rankings
Excels at
It's decent at
Falls behind in
Detailed scores
Updated: Aug 23, 2026Overall
undefined metric} other undefined metrics}}
Overall score68.0#24 / 100DeepSWE42.2#15 / 16
Overall
undefined metric} other undefined metrics}}
Benchmark
LMSpeed rank
Score
Source
Updated
Confidence
Pricing
undefined metric} other undefined metrics}}
Input price$0.425/M#79 / 180Output price$3.10/M#102 / 180
Pricing
undefined metric} other undefined metrics}}
Benchmark
LMSpeed rank
Score
Source
Updated
Confidence
Agents
V3.0undefined metric} other undefined metrics}} · Estimated
Score57.280% interval45.3–69.12/4 Measured dimensionsAgentic score84.4#12 / 69Terminal-Bench 2.173.0#6 / 7
Agents
V3.0undefined metric} other undefined metrics}} · Estimated
Dimensions and evidence
Planning & decompositionPrior only
Tool usePrior only
Environment & long-horizon execution62
undefined family} other undefined families}} · undefined metric} other undefined metrics}}
osworld · os_world_verified · z 1.24 · q 1.00
Recovery & completion reliability52.4
undefined family} other undefined families}} · undefined metric} other undefined metrics}}
jobbench · job_bench · z 0.07 · q 1.00
Benchmark
LMSpeed rank
Score
Source
Updated
Confidence
Coding
V3.0undefined metric} other undefined metrics}} · Estimated
Score56.280% interval45.6–66.92/4 Measured dimensionsSciCode44.7%#48 / 204Coding score53.3#51 / 81
Coding
V3.0undefined metric} other undefined metrics}} · Estimated
Dimensions and evidence
Code generation59.9
undefined family} other undefined families}} · undefined metric} other undefined metrics}}
livecodebench · live_code_bench_v6 · z 1.49 · q 0.75
scicode · scicode · z 0.59 · q 1.00
Repository engineering52.6
undefined family} other undefined families}} · undefined metric} other undefined metrics}}
nl2repo · nl2_repo · z -0.10 · q 1.00
swe_pro · swe_pro · z 0.91 · q 1.00
Debugging & testingPrior only
Tool-assisted development & qualityPrior only
Benchmark
LMSpeed rank
Score
Source
Updated
Confidence
SciCodeV3 evidence
LMSpeed rank#48
Score44.7%
Sourceartificialanalysis.ai
UpdatedAug 23, 2026
Confidence4
Reasoning
V3.0undefined metric} other undefined metrics}} · Provisional
Score58.180% interval44.2–72.01/4 Measured dimensionsGPQA90.5%#25 / 211HLE33.9%#40 / 208
Reasoning
V3.0undefined metric} other undefined metrics}} · Provisional
Dimensions and evidence
Abstract logicPrior only
Scientific & causal reasoning58.1
undefined family} other undefined families}} · undefined metric} other undefined metrics}}
gpqa · gpqa · z 0.98 · q 1.00
hle · hle · z 0.76 · q 1.00
hle · hle_no_tools · z -1.21 · q 1.00
Multi-step constraintsPrior only
Evidence integration & verificationPrior only
Benchmark
LMSpeed rank
Score
Source
Updated
Confidence
Knowledge
V3.0undefined metric} other undefined metrics}} · Provisional
Score41.580% interval25.4–57.51/4 Measured dimensionsKnowledge score47.0#62 / 69GPQA-D89.2#18 / 31
Knowledge
V3.0undefined metric} other undefined metrics}} · Provisional
Dimensions and evidence
Broad knowledge41.5
undefined family} other undefined families}} · undefined metric} other undefined metrics}}
benchlm_category_knowledge · benchlm_category_knowledge · z -1.47 · q 1.00
Professional knowledgePrior only
FactualityPrior only
Retrieval & open-book usePrior only
Benchmark
LMSpeed rank
Score
Source
Updated
Confidence
Math
V3.0undefined metric} other undefined metrics}} · No data
No data80% interval30.8–69.20/4 Measured dimensions
Math
V3.0undefined metric} other undefined metrics}} · No data
Dimensions and evidence
Foundational mathPrior only
Competition mathPrior only
Advanced proofsPrior only
Applied & tool-assisted mathPrior only
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}} · Estimated
Score57.780% interval46.4–69.02/4 Measured dimensionsMultimodal Grounded score79.8#9 / 44MathVision90.0#4 / 13
Multimodal
V3.0undefined metric} other undefined metrics}} · Estimated
Dimensions and evidence
Perception & OCRPrior only
Document & spatial understanding59.4
undefined family} other undefined families}} · undefined metric} other undefined metrics}}
charxiv · charxiv · z 1.37 · q 1.00
Visual reasoning56
undefined family} other undefined families}} · undefined metric} other undefined metrics}}
erqa · erqa · z 0.01 · q 1.00
mathvision · math_vision · z 0.92 · q 1.00
Video & grounded actionPrior only
Benchmark
LMSpeed rank
Score
Source
Updated
Confidence
Instruction following
V3.0undefined metric} other undefined metrics}} · Provisional
Score56.180% interval40.1–72.21/4 Measured dimensionsInstruction Following score90.4#10 / 25IFBench79.5#5 / 14
Instruction following
V3.0undefined metric} other undefined metrics}} · Provisional
Dimensions and evidence
Constraint followingPrior only
Structured outputPrior only
Novel-instruction generalization56.1
undefined family} other undefined families}} · undefined metric} other undefined metrics}}
ifbench · if_bench · z 0.55 · q 1.00
Multi-turn & long instructionsPrior only
Benchmark
LMSpeed rank
Score
Source
Updated
Confidence
OpenRouter endpoints
8 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 |
|---|---|---|---|---|---|---|
Alibaba alibaba | $0.500/M | $3/M | 100.0% | — | — | undefined tokens / undefined tokens |
AkashML akashml/bf16 | $0.400/M | $3/M | 99.9% | — | — | undefined tokens / undefined tokens |
Reka reka/fp8 | $0.450/M | $3.20/M | 99.9% | — | — | undefined tokens / undefined tokens |
Parasail parasail/fp8 | $0.450/M | $3.20/M | 99.6% | — | — | undefined tokens / undefined tokens |
Chutes chutes/fp8 | $0.400/M | $3/M | 99.5% | — | — | undefined tokens / undefined tokens |
Io Net io-net/fp8 | $0.480/M | $3.40/M | 99.3% | — | — | undefined tokens / undefined tokens |
Venice venice/fp8 | $0.450/M | $3.20/M | 99.3% | — | — | undefined tokens / undefined tokens |
CoreWeave coreweave/fp8 | $0.400/M | $3/M | 98.0% | — | — | undefined tokens / undefined tokens |