Agentic
Like-for-like- GPT-5.2
- 43.0
- Supported · #109/151
- Qwen3.6-27B
- 33.9
- Supported · #133/151
- Basis
- BenchAlign lane · 4 vs 6 public rows
- Reading
- GPT-5.2 leads · intervals overlap
GPT-5.2 vs Qwen3.6-27B
Updated September 4, 2026. Public scores include evidence status and uncertainty. They are not guarantees for a specific workload.
Decision reading
GPT-5.2 has the higher public score, 65.45 versus 52.73, and the 90% score intervals do not overlap.
7 results are shared. Category rows resting on Estimated evidence or different benchmark sets are marked directional and do not name a winner.
Share or export
Which one for your work
Recommendations appear only when a shared evidence basis or an explicit operating constraint supports the call. Secondary and unsupported use cases stay disclosed below the initial list.
Coding work
Code generation, repair, and software-engineering tasks
GPT-5.2
GPT-5.2 leads on the public coding lane, 47.2 to 42.6, with Supported evidence for both models, although the 90% intervals overlap.
Confidence: limited
Agentic work
Tool use, computer use, and multi-step task completion
GPT-5.2
GPT-5.2 leads on the public agentic lane, 43 to 33.9, with Supported evidence for both models, although the 90% intervals overlap.
Confidence: limited
Long documents
Prompts that approach the documented context limit
GPT-5.2
GPT-5.2 has the larger documented context window.
Confidence: documented
Show secondary and unsupported calls
Chat turn cost
1K fresh input + 500 output tokens
Not enough matched evidence
A complete comparable API-rate estimate is not available for both models.
Confidence: listed-rates
Cache-heavy agent loop cost
200K cached + 20K fresh input + 10K output tokens
Not enough matched evidence
A complete comparable API-rate estimate is not available for both models.
Confidence: rate-fallback
Repository review cost
50K fresh input + 3K output tokens
Not enough matched evidence
A complete comparable API-rate estimate is not available for both models.
Confidence: listed-rates
What is actually comparable
Shared results can support a head-to-head reading. Results present for only one model describe coverage, not superiority.
8 GPT-5.2 only
7 Shared
31 Qwen3.6-27B only
- Shared results
- 7
- GPT-5.2 only
- 8
- Qwen3.6-27B only
- 31
- Like-for-like categories
- 3 / 8
2 categories rest on Estimated evidence or different evidence sets. Those rows remain visible for coverage context but do not name a winner.
Category results, on a stated basis
Each row shows the public-lane category score for both models: the BenchAlign lane for agentic, coding, and knowledge, and the weighted public rows elsewhere, with the same rank each model page reports. A row is like-for-like only when both scores rest on Supported evidence or the same weighted set. Directional and not-comparable rows remain visible, but they never receive a winner.
Coding
Like-for-like- GPT-5.2
- 47.2
- Supported · #91/183
- Qwen3.6-27B
- 42.6
- Supported · #128/183
- Basis
- BenchAlign lane · 3 vs 6 public rows
- Reading
- GPT-5.2 leads · intervals overlap
Multimodal
Like-for-like- GPT-5.2
- 65.6
- #23/48
- Qwen3.6-27B
- 51.5
- #35/48
- Basis
- Provisional lane · 2 vs 2 weighted rows
- Reading
- GPT-5.2 leads
Knowledge
Directional only- GPT-5.2
- 62.0
- Supported · #30/181
- Qwen3.6-27B
- 49.0
- Estimated · #95/181
- Basis
- BenchAlign lane · 1 vs 6 public rows
- Reading
- Directional only
Instruction following
Directional only- GPT-5.2
- 92.3
- #14/120
- Qwen3.6-27B
- 82.2
- #50/120
- Basis
- Provisional lane · 0 vs 0 weighted rows
- Reading
- Directional only
Reasoning
Not comparable- GPT-5.2
- 53.8
- Unranked · 3 rankable rows
- Qwen3.6-27B
- 73.7
- Unranked · 2 rankable rows
- Basis
- Provisional lane · 1 vs 0 weighted rows
- Reading
- Not comparable
Math
Not comparable- GPT-5.2
- 57.5
- Unranked · 2 rankable rows
- Qwen3.6-27B
- 72.8
- Unranked · 5 rankable rows
- Basis
- Provisional lane · 2 vs 2 weighted rows
- Reading
- Not comparable
Multilingual
Not comparable- GPT-5.2
- Not ranked
- Qwen3.6-27B
- Not ranked
- Basis
- Provisional lane · 0 vs 0 weighted rows
- Reading
- Not comparable
| Category | GPT-5.2 | Qwen3.6-27B | Basis | Reading |
|---|---|---|---|---|
| Agentic | 43.0Supported · #109/151 | 33.9Supported · #133/151 | Like-for-likeBenchAlign lane · 4 vs 6 public rows | GPT-5.2 leads · intervals overlap |
| Coding | 47.2Supported · #91/183 | 42.6Supported · #128/183 | Like-for-likeBenchAlign lane · 3 vs 6 public rows | GPT-5.2 leads · intervals overlap |
| Multimodal | 65.6#23/48 | 51.5#35/48 | Like-for-likeProvisional lane · 2 vs 2 weighted rows | GPT-5.2 leads |
| Knowledge | 62.0Supported · #30/181 | 49.0Estimated · #95/181 | Directional onlyBenchAlign lane · 1 vs 6 public rows | Directional only |
| Instruction following | 92.3#14/120 | 82.2#50/120 | Directional onlyProvisional lane · 0 vs 0 weighted rows | Directional only |
| Reasoning | 53.8Unranked · 3 rankable rows | 73.7Unranked · 2 rankable rows | Not comparableProvisional lane · 1 vs 0 weighted rows | Not comparable |
| Math | 57.5Unranked · 2 rankable rows | 72.8Unranked · 5 rankable rows | Not comparableProvisional lane · 2 vs 2 weighted rows | Not comparable |
| Multilingual | Not ranked | Not ranked | Not comparableProvisional lane · 0 vs 0 weighted rows | Not comparable |
Ranks count the models scored in each category’s lane, so the agentic, coding, and knowledge denominators (BenchAlign) differ from the provisional-lane categories. Unranked scores sit on the lane’s scale but never name a winner. Like-for-like rows within 0.5 points read as a practical tie.
Shape of the matched evidence
Only shared public evidence is shown. Sparse evidence stays a ruled list rather than being closed into a radar shape.
Too few matched category axes support a radar. The ruled list below shows only shared benchmark results; positions use each benchmark’s normalized display scale when available.
GPQA
Knowledge
MMMU-Pro
Multimodal
CharXiv
Multimodal
SWE-bench Verified
Coding
SWE-bench Pro
Coding
What each workload costs
Three fixed token mixes turn per-token rates into comparable decisions. Each scenario states context fit and whether cached input had to fall back to the published list-input rate.
Chat turn
1K fresh input + 500 output tokens
- GPT-5.2
- $0.00875
- Fits in one request
- Qwen3.6-27B
- Self-hosted; infrastructure cost varies
- Fits in one request
Qwen3.6-27B has no comparable published API token rate.
Repository review
50K fresh input + 3K output tokens
- GPT-5.2
- $0.1295
- Fits in one request
- Qwen3.6-27B
- Self-hosted; infrastructure cost varies
- Fits in one request
Qwen3.6-27B has no comparable published API token rate.
Cache-heavy agent loop
200K cached + 20K fresh input + 10K output tokens
- GPT-5.2
- $0.525
- Fits in one request
- Cached input priced at the published list-input rate
- Qwen3.6-27B
- Self-hosted; infrastructure cost varies
- Fits in one request
- Cached-input rate unavailable
GPT-5.2 has no published cached-input rate, so cached tokens use its listed input rate. Qwen3.6-27B has no comparable published API token rate.
Specification differences
Sourced differences are shown directly. Missing facts stay explicit instead of being inferred from a model name or family.
SpecificationGPT-5.2Qwen3.6-27B
Context window
Maximum documented context; output-token limits may be lower.
GPT-5.2
400K
Qwen3.6-27B
262K
API model ID
GPT-5.2
Not sourced
Qwen3.6-27B
Not sourced
Cached-input rate
A missing cached-input rate falls back to the listed input rate only in the stated workload estimate.
GPT-5.2
Not published
Qwen3.6-27B
No comparable hosted API rate
Documented inputs
GPT-5.2
Not sourced
Qwen3.6-27B
Not sourced
Documented outputs
GPT-5.2
Not sourced
Qwen3.6-27B
Not sourced
Provider availability
GPT-5.2
Not sourced
Qwen3.6-27B
Not sourced
Reasoning profile
GPT-5.2
Reasoning
Qwen3.6-27B
Reasoning
Weight access
GPT-5.2
Proprietary
Qwen3.6-27B
Open Weight
License
GPT-5.2
Proprietary
Qwen3.6-27B
Open Weight
Release date
GPT-5.2
2025-12-11
Qwen3.6-27B
2026-04-21
If you already use one of these models
- Deployment change
- The models list different providers, so authentication, endpoint behavior, limits, and feature support may change.
- Quality signal
- GPT-5.2 has the higher public score, 65.45 versus 52.73, and the 90% score intervals do not overlap.
- Workload cost
- A complete comparable API-rate estimate is not available for both models.
- Context tradeoff
- GPT-5.2 has the larger documented window (400K).
Run the same representative tasks against both endpoints before changing production traffic.
Self-host vs API cost
Estimates at 50,000 req/day · 1000 tokens/req average.
GPT-5.2
API / mo$11,813
Self-host / moNot listed
Break-even—
Proprietary model — self-hosting not applicable.
Qwen3.6-27B
API / mo$0
Self-host / mo$429
Break-even—
Benchmark evidence
The full public result ledger is available for audit without forcing a wide desktop table onto a phone.
Browse raw public benchmark evidence46 rows
Agentic
BrowseComp
GPT-5.265.8%SourceQwen3.6-27B—Not directly comparable
OSWorld-Verified
GPT-5.247.3%SourceQwen3.6-27B—Not directly comparable
Gert Labs
Shared sourceGPT-5.246.54%Qwen3.6-27B54.84%Qwen3.6-27B leads this result
JobBench
GPT-5.234.3%SourceQwen3.6-27B—Not directly comparable
Terminal-Bench 2.0
GPT-5.2—Qwen3.6-27B59.3%SourceNot directly comparable
Claw-Eval
GPT-5.2—Qwen3.6-27B72.4%SourceNot directly comparable
QwenClawBench
GPT-5.2—Qwen3.6-27B53.4%SourceNot directly comparable
QwenWebBench
GPT-5.2—Qwen3.6-27B1487SourceNot directly comparable
AndroidWorld
GPT-5.2—Qwen3.6-27B70.3%SourceNot directly comparable
Coding
SWE-bench Verified
GPT-5.280%SourceQwen3.6-27B77.2%SourceGPT-5.2 leads this result
SWE-bench Pro
GPT-5.255.6%SourceQwen3.6-27B53.5%SourceGPT-5.2 leads this result
Vibe Code Bench
GPT-5.253.50%SourceQwen3.6-27B—Not directly comparable
SWE Multilingual
GPT-5.2—Qwen3.6-27B71.3%SourceNot directly comparable
Terminal-Bench 2.0
GPT-5.2—Qwen3.6-27B59.3%SourceNot directly comparable
LiveCodeBench
GPT-5.2—Qwen3.6-27B83.9%SourceNot directly comparable
NL2Repo
GPT-5.2—Qwen3.6-27B36.2%SourceNot directly comparable
Reasoning
ARC-AGI-2
GPT-5.252.9%SourceQwen3.6-27B—Not directly comparable
Knowledge
GPQA
GPT-5.292.4%SourceQwen3.6-27B87.8%SourceGPT-5.2 leads this result
MMLU-Pro
GPT-5.2—Qwen3.6-27B86.2%SourceNot directly comparable
MMLU-Redux
GPT-5.2—Qwen3.6-27B93.5%SourceNot directly comparable
SuperGPQA
GPT-5.2—Qwen3.6-27B66%SourceNot directly comparable
C-Eval
GPT-5.2—Qwen3.6-27B91.4%SourceNot directly comparable
HLE
GPT-5.2—Qwen3.6-27B24%SourceNot directly comparable
Math
FrontierMath v2 (Tiers 1-3)
GPT-5.240.700%SourceQwen3.6-27B—Not directly comparable
FrontierMath v2 (Tier 4)
GPT-5.218.800%SourceQwen3.6-27B—Not directly comparable
HMMT Feb 2025
GPT-5.2—Qwen3.6-27B93.8%SourceNot directly comparable
HMMT Nov 2025
GPT-5.2—Qwen3.6-27B90.7%SourceNot directly comparable
HMMT Feb 2026
GPT-5.2—Qwen3.6-27B84.3%SourceNot directly comparable
MMAnswerBench
GPT-5.2—Qwen3.6-27B80.8%SourceNot directly comparable
AIME26
GPT-5.2—Qwen3.6-27B94.1%SourceNot directly comparable
Multimodal
MMMU-Pro
GPT-5.279.5%SourceQwen3.6-27B75.8%SourceGPT-5.2 leads this result
MathVision
GPT-5.283.0%SourceQwen3.6-27B—Not directly comparable
CharXiv
GPT-5.282.1%SourceQwen3.6-27B78.4%SourceGPT-5.2 leads this result
V*
GPT-5.275.9%SourceQwen3.6-27B94.7%SourceQwen3.6-27B leads this result
MMMU
GPT-5.2—Qwen3.6-27B82.9%SourceNot directly comparable
RealWorldQA
GPT-5.2—Qwen3.6-27B84.1%SourceNot directly comparable
DynaMath
GPT-5.2—Qwen3.6-27B85.6%SourceNot directly comparable
MStar
GPT-5.2—Qwen3.6-27B81.4%SourceNot directly comparable
SimpleVQA
GPT-5.2—Qwen3.6-27B56.1%SourceNot directly comparable
CC-OCR
GPT-5.2—Qwen3.6-27B81.2%SourceNot directly comparable
CountBench
GPT-5.2—Qwen3.6-27B97.8%SourceNot directly comparable
RefCOCO (avg)
GPT-5.2—Qwen3.6-27B92.5%SourceNot directly comparable
ERQA
GPT-5.2—Qwen3.6-27B62.5%SourceNot directly comparable
Video-MME (with subtitle)
GPT-5.2—Qwen3.6-27B87.7%SourceNot directly comparable
VideoMMMU
GPT-5.2—Qwen3.6-27B84.4%SourceNot directly comparable
MLVU (M-Avg)
GPT-5.2—Qwen3.6-27B86.6%SourceNot directly comparable
Frequently asked questions
Which is better, GPT-5.2 or Qwen3.6-27B?
GPT-5.2 has the higher public score, 65.45 versus 52.73, and the 90% score intervals do not overlap. That is the clearest overall quality signal in the current public data.
Which is better for coding, GPT-5.2 or Qwen3.6-27B?
GPT-5.2 leads the public coding lane, 47.2 to 42.6, with Supported evidence for both models, although the 90% intervals overlap.
Which is better for agentic tasks, GPT-5.2 or Qwen3.6-27B?
GPT-5.2 leads the public agentic tasks lane, 43 to 33.9, with Supported evidence for both models, although the 90% intervals overlap.
Which costs less, GPT-5.2 or Qwen3.6-27B?
Both models do not have comparable published API token rates, so this page does not name a universal price winner.
Which has the larger context window, GPT-5.2 or Qwen3.6-27B?
GPT-5.2 has the larger documented context window: 400K, compared with 262K.
Related comparisons
Last updated September 4, 2026
Watch GPT-5.2 vs Qwen3.6-27B
One weekly email when material rank, price, or benchmark evidence changes make this matchup worth revisiting.
Read a sample issueJoin 2,000+ readers.