Long documents
Prompts that approach the documented context limit
Hy4 preview
Hy4 preview has the larger documented context window.
Confidence: documented
Hy4 preview vs Qwen3.5 397B
Updated October 13, 2026. Rank says Hy4 preview is ahead. Price, access, and your workload can each overturn that. Public scores include evidence status and uncertainty.
Both of these models will change. Get the price, version and retirement notices for the pair, each with its source. Follow model changes
Share or export
Decision reading
Hy4 preview has the higher public point estimate, 60.86 versus 55.62. Their conditional score ranges overlap. These ranges do not establish rank confidence. 7 results are shared. Category rows resting on Estimated evidence or different benchmark sets are marked directional and do not name a winner.
- Shared results
- 7
- Hy4 preview only
- 23
- Qwen3.5 397B only
- 55
- Like-for-like categories
- 0 / 8
Estimated: Hy4 preview and Qwen3.5 397B. Conditional ranges do not establish rank confidence.How the comparison works
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.
Show secondary and unsupported calls
Coding work
Code generation, repair, and software-engineering tasks
Not enough matched evidence
Qwen3.5 397B is not ranked on the public lane for coding, so no winner is named for coding.
Confidence: limitedAgentic work
Tool use, computer use, and multi-step task completion
Not enough matched evidence
Hy4 preview and Qwen3.5 397B are scored on Estimated evidence for agentic, so the reading is directional rather than like-for-like.
Confidence: limitedChat 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-ratesCache-heavy agent loop cost
200K cached + 20K fresh input + 10K output tokens
Not enough matched evidence
The page does not recommend a cost winner because at least one model cannot fit the stated workload in one request. Qwen3.5 397B does not fit this workload in one request. Qwen3.5 397B has no published cached-input rate, so cached tokens use its listed input rate. Hy4 preview has no comparable published API token rate.
Confidence: rate-fallbackRepository 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
Which one for a specific job
Choose a job from the LLM Selector's task catalog to see the category row it rests on, under the same basis rules as the table below. A directional row stays directional; choosing a job never creates a winner.
The same task catalog as the LLM Selector. Each job names the evidence surface it rests on; nothing here adds a new score.
57.1Hy4 preview—Qwen3.5 397B
Not comparable · BenchAlign v5.8
The coding row is not comparable on the public lane: at least one model is not measured or not ranked there.
Coding scores combine specific tasks and setups. Match the editor, harness, and effort to your workflow.
Same basis rules as the category table below
What is actually comparable
Shared results can support a head-to-head reading. Results present for only one model describe coverage, not superiority.
23 Hy4 preview only
7 Shared
55 Qwen3.5 397B only
2 categories rest on Estimated evidence or different evidence sets. Those rows remain visible for coverage context but do not name a winner.
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.
HLEKnowledge
Normalized gap 26.7Hy4 preview:55.4%Qwen3.5 397B:28.7%SWE-bench ProCoding
Normalized gap 14.8Hy4 preview:65.7%Qwen3.5 397B:50.9%SWE MultilingualCoding
Normalized gap 13.6Hy4 preview:82.9%Qwen3.5 397B:69.3%GPQAKnowledge
Normalized gap 3.9Hy4 preview:92.3%Qwen3.5 397B:88.4%
Bars run 0–100 on each benchmark’s normalized display scale
Category results, on a stated basis
Each row shows the public-lane category score for both models: the BenchAlign v5.8 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.
Agentic
Directional only- Hy4 preview
- 54.5
- Estimated · #37/123
- Qwen3.5 397B
- 44.3
- Estimated · #55/123
- Basis
- BenchAlign v5.8 lane · 15 vs 14 public rows
- Reading
- Directional only
Knowledge
Directional only- Hy4 preview
- 63.6
- Estimated · #39/180
- Qwen3.5 397B
- 53.8
- Estimated · #64/180
- Basis
- BenchAlign v5.8 lane · 4 vs 6 public rows
- Reading
- Directional only
Coding
Not comparable- Hy4 preview
- 57.1
- Supported · #26/148
- Qwen3.5 397B
- Not ranked
- Basis
- BenchAlign v5.8 lane · 8 vs 6 public rows
- Reading
- Not comparable
Reasoning
Not comparable- Hy4 preview
- Not ranked
- Qwen3.5 397B
- 74.9
- Unranked · 2 rankable rows
- Basis
- Provisional lane · 0 vs 1 weighted rows
- Reading
- Not comparable
Multimodal
Not comparable- Hy4 preview
- 88.3
- Unranked · 1 rankable row
- Qwen3.5 397B
- 73.6
- #29/54
- Basis
- Provisional lane · 1 vs 2 weighted rows
- Reading
- Not comparable
Multilingual
Not comparable- Hy4 preview
- Not ranked
- Qwen3.5 397B
- 93.8
- #5/17
- Basis
- Provisional lane · 0 vs 1 weighted rows
- Reading
- Not comparable
Instruction following
Not comparable- Hy4 preview
- Not ranked
- Qwen3.5 397B
- 79.9
- #56/127
- Basis
- Provisional lane · 0 vs 1 weighted rows
- Reading
- Not comparable
Math
Not comparable- Hy4 preview
- Not ranked
- Qwen3.5 397B
- 81.7
- Unranked · 5 rankable rows
- Basis
- Provisional lane · 0 vs 2 weighted rows
- Reading
- Not comparable
| Category | Hy4 preview | Qwen3.5 397B | Basis | Reading |
|---|---|---|---|---|
| Agentic | 54.5Estimated · #37/123 | 44.3Estimated · #55/123 | Directional onlyBenchAlign v5.8 lane · 15 vs 14 public rows | Directional only |
| Knowledge | 63.6Estimated · #39/180 | 53.8Estimated · #64/180 | Directional onlyBenchAlign v5.8 lane · 4 vs 6 public rows | Directional only |
| Coding | 57.1Supported · #26/148 | Not ranked | Not comparableBenchAlign v5.8 lane · 8 vs 6 public rows | Not comparable |
| Reasoning | Not ranked | 74.9Unranked · 2 rankable rows | Not comparableProvisional lane · 0 vs 1 weighted rows | Not comparable |
| Multimodal | 88.3Unranked · 1 rankable row | 73.6#29/54 | Not comparableProvisional lane · 1 vs 2 weighted rows | Not comparable |
| Multilingual | Not ranked | 93.8#5/17 | Not comparableProvisional lane · 0 vs 1 weighted rows | Not comparable |
| Instruction following | Not ranked | 79.9#56/127 | Not comparableProvisional lane · 0 vs 1 weighted rows | Not comparable |
| Math | Not ranked | 81.7Unranked · 5 rankable rows | Not comparableProvisional lane · 0 vs 2 weighted rows | Not comparable |
Ranks count the models scored in each category’s lane, so the agentic, coding, and knowledge denominators (BenchAlign v5.8) differ from the provisional-lane categories. Unranked scores sit on the provisional lane’s scale but never name a winner. Like-for-like rows within 0.5 points read as a practical tie.
Supported evidence per lane · bars run 0–100Methodology
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
- Hy4 preview
- Self-hosted; infrastructure cost varies
- Fits in one request
- Qwen3.5 397B
- $0.0024
- Fits in one request
Hy4 preview has no comparable published API token rate.
Repository review
50K fresh input + 3K output tokens
- Hy4 preview
- Self-hosted; infrastructure cost varies
- Fits in one request
- Qwen3.5 397B
- $0.0408
- Fits in one request
Hy4 preview has no comparable published API token rate.
Cache-heavy agent loop
200K cached + 20K fresh input + 10K output tokens
- Hy4 preview
- Self-hosted; infrastructure cost varies
- Fits in one request
- Cached-input rate unavailable
- Qwen3.5 397B
- $0.168
- Does not fit in one request
- Cached input priced at the published list-input rate
Qwen3.5 397B does not fit this workload in one request. Qwen3.5 397B has no published cached-input rate, so cached tokens use its listed input rate. Hy4 preview has no comparable published API token rate.
Cached input falls back to the list input rate only where a cached rate is unpublished
Specification differences
Sourced differences are shown directly. Missing facts stay explicit instead of being inferred from a model name or family.
SpecificationHy4 previewQwen3.5 397B
Context window
Maximum documented context; output-token limits may be lower.
Hy4 preview
Qwen3.5 397B
128K
API model ID
Hy4 preview
tencent/Hy4-preview
Tencent Hy4 preview model cardQwen3.5 397B
Not sourced
Cached-input rate
A missing cached-input rate falls back to the listed input rate only in the stated workload estimate.
Hy4 preview
No comparable hosted API rate
Tencent Hy4 preview model cardQwen3.5 397B
Not published
Documented inputs
Hy4 preview
Not sourced
Qwen3.5 397B
Not sourced
Documented outputs
Hy4 preview
Not sourced
Qwen3.5 397B
Not sourced
Provider availability
Hy4 preview
Not sourced
Qwen3.5 397B
Not sourced
Reasoning profile
Hy4 preview
Reasoning
Qwen3.5 397B
Non-Reasoning
Weight access
Hy4 preview
Open Weight
Qwen3.5 397B
Open Weight
License
Hy4 preview
Open Weight
Qwen3.5 397B
Open Weight
Release date
Hy4 preview
2026-08-28
Qwen3.5 397B
2026-02-16
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
- Hy4 preview has the higher public point estimate, 60.86 versus 55.62. Their conditional score ranges overlap. These ranges do not establish rank confidence.
- Workload cost
- A complete comparable API-rate estimate is not available for both models.
- Context tradeoff
- Hy4 preview has the larger documented window (1M).
Run the same representative tasks against both endpoints before changing production traffic.
Questions
Which is better, Hy4 preview or Qwen3.5 397B?
Hy4 preview has the higher public point estimate, 60.86 versus 55.62. Their conditional score ranges overlap. These ranges do not establish rank confidence. The page therefore keeps the decision tied to the specific documented workload.
Which is better for coding, Hy4 preview or Qwen3.5 397B?
Qwen3.5 397B is not ranked on the public lane for coding, so no winner is named for coding.
Which is better for agentic tasks, Hy4 preview or Qwen3.5 397B?
Hy4 preview scores higher for agentic tasks on the public lane, 54.5 to 44.3. Hy4 preview and Qwen3.5 397B are scored on Estimated evidence for agentic tasks, so the reading is directional rather than like-for-like. BenchLM does not name a winner for agentic tasks from a directional reading; read the shared benchmark rows directly and test the models on the same task set.
Which costs less, Hy4 preview or Qwen3.5 397B?
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, Hy4 preview or Qwen3.5 397B?
Hy4 preview has the larger documented context window: 1M, compared with 128K.
Benchmark evidence
The full public result ledger is available for audit without forcing a wide desktop table onto a phone.
Browse raw public benchmark evidence85 rows
Agentic
Terminal-Bench 2.1
Hy4 preview85.4%SourceQwen3.5 397B—Not directly comparable
CyberGym
Hy4 preview78.4%SourceQwen3.5 397B—Not directly comparable
WideResearch
Hy4 preview83.9%SourceQwen3.5 397B74.0%SourceHy4 preview leads this result
DRACO
Hy4 preview77.2%SourceQwen3.5 397B—Not directly comparable
MCP Atlas
Hy4 preview83.7%SourceQwen3.5 397B46.1%SourceHy4 preview leads this result
Toolathlon-Verified
Hy4 preview74.1%SourceQwen3.5 397B—Not directly comparable
APEX-Agents
Hy4 preview37.1%SourceQwen3.5 397B—Not directly comparable
skillsBench
Hy4 preview62.9%SourceQwen3.5 397B—Not directly comparable
JobBench
Hy4 preview61.7%SourceQwen3.5 397B—Not directly comparable
Agents' Last Exam
Hy4 preview22.8%SourceQwen3.5 397B—Not directly comparable
AutomationBench
Hy4 preview32.1%SourceQwen3.5 397B—Not directly comparable
BankerToolBench
Hy4 preview78.6%SourceQwen3.5 397B—Not directly comparable
HLE w/ tools
Hy4 preview55.4%SourceQwen3.5 397B—Not directly comparable
Terminal-Bench 2.1 (Vals)
Hy4 preview55.1%SourceQwen3.5 397B—Not directly comparable
CWE-bench v1
Hy4 preview53.0%SourceQwen3.5 397B—Not directly comparable
Terminal-Bench 2.0
Hy4 preview—Qwen3.5 397B52.5%SourceNot directly comparable
BrowseComp
Hy4 preview—Qwen3.5 397B69%SourceNot directly comparable
OSWorld-Verified
Hy4 preview—Qwen3.5 397B62.2%SourceNot directly comparable
Claw-Eval
Hy4 preview—Qwen3.5 397B56.8%SourceNot directly comparable
QwenClawBench
Hy4 preview—Qwen3.5 397B51.8%SourceNot directly comparable
τ³-bench results
Hy4 preview—Qwen3.5 397B68.4%SourceNot directly comparable
VITA-Bench
Hy4 preview—Qwen3.5 397B43.7%SourceNot directly comparable
DeepPlanning
Hy4 preview—Qwen3.5 397B37.6%SourceNot directly comparable
Toolathlon
Hy4 preview—Qwen3.5 397B36.3%SourceNot directly comparable
MCP-Tasks
Hy4 preview—Qwen3.5 397B74.2%SourceNot directly comparable
Gert Labs
Hy4 preview—Qwen3.5 397B46.76%SourceNot directly comparable
ResearchClawBench
Hy4 preview—Qwen3.5 397B14.2%SourceNot directly comparable
Coding
Terminal-Bench 2.1
Hy4 preview85.4%SourceQwen3.5 397B—Not directly comparable
SWE-bench Pro
Hy4 preview65.7%SourceQwen3.5 397B50.9%SourceHy4 preview leads this result
SWE Multilingual
Hy4 preview82.9%SourceQwen3.5 397B69.3%SourceHy4 preview leads this result
DeepSWE
Hy4 preview64.3%SourceQwen3.5 397B—Not directly comparable
NL2Repo
Hy4 preview58.9%SourceQwen3.5 397B32.2%SourceHy4 preview leads this result
ProgramBench
Hy4 preview17.5%SourceQwen3.5 397B—Not directly comparable
PostTrain Bench
Hy4 preview35.6%SourceQwen3.5 397B—Not directly comparable
sweMarathon
Hy4 preview31.9%SourceQwen3.5 397B—Not directly comparable
SWE-bench Verified
Hy4 preview—Qwen3.5 397B76.2%SourceNot directly comparable
LiveCodeBench v6
Hy4 preview—Qwen3.5 397B83.6%SourceNot directly comparable
Terminal-Bench 2.0
Hy4 preview—Qwen3.5 397B52.5%SourceNot directly comparable
Reasoning
Multimodal
OfficeQA Pro
Hy4 preview66.2%SourceQwen3.5 397B—Not directly comparable
MMMU
Hy4 preview—Qwen3.5 397B85.0%SourceNot directly comparable
MMMU-Pro
Hy4 preview—Qwen3.5 397B79%SourceNot directly comparable
RealWorldQA
Hy4 preview—Qwen3.5 397B83.9%SourceNot directly comparable
OmniDocBench 1.5
Hy4 preview—Qwen3.5 397B90.8%SourceNot directly comparable
Video-MME (with subtitle)
Hy4 preview—Qwen3.5 397B87.5%SourceNot directly comparable
Video-MME (w/o subtitle)
Hy4 preview—Qwen3.5 397B83.7%SourceNot directly comparable
MathVision
Hy4 preview—Qwen3.5 397B88.6%SourceNot directly comparable
We-Math
Hy4 preview—Qwen3.5 397B87.9%SourceNot directly comparable
DynaMath
Hy4 preview—Qwen3.5 397B86.3%SourceNot directly comparable
SimpleVQA
Hy4 preview—Qwen3.5 397B67.1%SourceNot directly comparable
CharXiv
Hy4 preview—Qwen3.5 397B80.8%SourceNot directly comparable
MMLongBench-Doc
Hy4 preview—Qwen3.5 397B61.5%SourceNot directly comparable
CC-OCR
Hy4 preview—Qwen3.5 397B82.0%SourceNot directly comparable
AI2D_TEST
Hy4 preview—Qwen3.5 397B93.9%SourceNot directly comparable
CountBench
Hy4 preview—Qwen3.5 397B97.2%SourceNot directly comparable
RefCOCO (avg)
Hy4 preview—Qwen3.5 397B92.3%SourceNot directly comparable
ODINW13
Hy4 preview—Qwen3.5 397B47.0%SourceNot directly comparable
ERQA
Hy4 preview—Qwen3.5 397B67.5%SourceNot directly comparable
VideoMMMU
Hy4 preview—Qwen3.5 397B84.7%SourceNot directly comparable
MLVU (M-Avg)
Hy4 preview—Qwen3.5 397B86.7%SourceNot directly comparable
ScreenSpot Pro
Hy4 preview—Qwen3.5 397B65.6%SourceNot directly comparable
V*
Hy4 preview—Qwen3.5 397B95.8%SourceNot directly comparable
Knowledge
GPQA
Hy4 preview92.3%SourceQwen3.5 397B88.4%SourceHy4 preview leads this result
GPQA-D
Hy4 preview92.3%SourceQwen3.5 397B—Not directly comparable
HLE
Hy4 preview55.4%SourceQwen3.5 397B28.7%SourceHy4 preview leads this result
HLE w/o tools
Hy4 preview43.4%SourceQwen3.5 397B—Not directly comparable
SuperGPQA
Hy4 preview—Qwen3.5 397B70.4%SourceNot directly comparable
MMLU-Pro
Hy4 preview—Qwen3.5 397B87.8%SourceNot directly comparable
MMLU-Redux
Hy4 preview—Qwen3.5 397B94.9%SourceNot directly comparable
C-Eval
Hy4 preview—Qwen3.5 397B93%SourceNot directly comparable
Multilingual
MMLU-ProX
Hy4 preview—Qwen3.5 397B84.7%SourceNot directly comparable
NOVA-63
Hy4 preview—Qwen3.5 397B59.1%SourceNot directly comparable
INCLUDE
Hy4 preview—Qwen3.5 397B85.6%SourceNot directly comparable
PolyMath
Hy4 preview—Qwen3.5 397B73.3%SourceNot directly comparable
MAXIFE
Hy4 preview—Qwen3.5 397B88.2%SourceNot directly comparable
Instruction following
Math
Apex
Hy4 preview74.2%SourceQwen3.5 397B—Not directly comparable
AIME26
Hy4 preview—Qwen3.5 397B93.3%SourceNot directly comparable
HMMT Feb 2025
Hy4 preview—Qwen3.5 397B94.8%SourceNot directly comparable
HMMT Nov 2025
Hy4 preview—Qwen3.5 397B92.7%SourceNot directly comparable
HMMT Feb 2026
Hy4 preview—Qwen3.5 397B87.9%SourceNot directly comparable
MMAnswerBench
Hy4 preview—Qwen3.5 397B80.9%SourceNot directly comparable
85 public results · 7 shared
Watch Hy4 preview vs Qwen3.5 397B
One weekly email when material rank, price, or benchmark evidence changes make this matchup worth revisiting.
Read a sample issueJoin 5,500+ readers.
Last updated October 13, 2026