Coding
Directional only- Granite 4.2 8B
- 36.5
- Qwen3.5 397B
- 66.5
- Weighted basis
- 3 vs 2 rows
- Reading
- Directional only
Granite 4.2 8B vs Qwen3.5 397B
Updated August 31, 2026. Public scores include evidence status and uncertainty. They are not guarantees for a specific workload.
Decision reading
Qwen3.5 397B has the higher public score estimate, 58.2 versus 46.88, but the 90% score intervals overlap. Treat that as a lead, not a settled winner.
7 results are shared. Category rows based on different benchmark sets are marked directional and do not name a winner.
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.
No workload recommendation clears the current evidence threshold.
Use the matched evidence, workload costs, and sourced specifications below instead of treating a point score as a universal answer.
Show secondary and unsupported calls
Coding work
Code generation, repair, and software-engineering tasks
Not enough matched evidence
The category averages use different weighted benchmark sets, so they are directional rather than like-for-like.
Confidence: limited
Agentic work
Tool use, computer use, and multi-step task completion
Not enough matched evidence
No shared weighted benchmark basis supports a winner.
Confidence: limited
Long documents
Prompts that approach the documented context limit
No clear pick
The documented context windows are equal.
Confidence: documented
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
The page does not recommend a cost winner because at least one model cannot fit the stated workload in one request. Granite 4.2 8B does not fit this 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. Granite 4.2 8B has no comparable published API token rate.
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.
7 Granite 4.2 8B only
7 Shared
31 Qwen3.5 397B only
- Shared results
- 7
- Granite 4.2 8B only
- 7
- Qwen3.5 397B only
- 31
- Like-for-like categories
- 0 / 8
2 categories use different evidence sets. Those rows remain visible for coverage context but do not name a winner.
Category results, on a stated basis
Each row states whether both averages use the same weighted benchmark set. Directional and not-comparable rows remain visible, but they never receive a winner in this template.
Knowledge
Directional only- Granite 4.2 8B
- 72.2
- Qwen3.5 397B
- 56.6
- Weighted basis
- 2 vs 4 rows
- Reading
- Directional only
Agentic
Not comparable- Granite 4.2 8B
- Not measured
- Qwen3.5 397B
- 56.5
- Weighted basis
- 0 vs 2 rows
- Reading
- Not comparable
Reasoning
Not comparable- Granite 4.2 8B
- Not measured
- Qwen3.5 397B
- 63.2
- Weighted basis
- 0 vs 1 rows
- Reading
- Not comparable
Math
Not comparable- Granite 4.2 8B
- Not measured
- Qwen3.5 397B
- 90.6
- Weighted basis
- 0 vs 2 rows
- Reading
- Not comparable
Multilingual
Not comparable- Granite 4.2 8B
- Not measured
- Qwen3.5 397B
- 84.7
- Weighted basis
- 0 vs 1 rows
- Reading
- Not comparable
Multimodal
Not comparable- Granite 4.2 8B
- Not measured
- Qwen3.5 397B
- 79.6
- Weighted basis
- 0 vs 2 rows
- Reading
- Not comparable
Instruction following
Not comparable- Granite 4.2 8B
- 79.3
- Qwen3.5 397B
- 92.6
- Weighted basis
- 1 vs 1 rows
- Reading
- Not comparable
| Category | Granite 4.2 8B | Qwen3.5 397B | Weighted basis | Reading |
|---|---|---|---|---|
| Coding | 36.5 | 66.5 | Directional only3 vs 2 rows | Directional only |
| Knowledge | 72.2 | 56.6 | Directional only2 vs 4 rows | Directional only |
| Agentic | Not measured | 56.5 | Not comparable0 vs 2 rows | Not comparable |
| Reasoning | Not measured | 63.2 | Not comparable0 vs 1 rows | Not comparable |
| Math | Not measured | 90.6 | Not comparable0 vs 2 rows | Not comparable |
| Multilingual | Not measured | 84.7 | Not comparable0 vs 1 rows | Not comparable |
| Multimodal | Not measured | 79.6 | Not comparable0 vs 2 rows | Not comparable |
| Instruction following | 79.3 | 92.6 | Not comparable1 vs 1 rows | Not comparable |
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.
SWE-bench Pro
Coding
Granite 4.2 8B: 19.1%Qwen3.5 397B: 50.9%Normalized gap 31.8Granite 4.2 8B source Qwen3.5 397B sourceSWE-bench Verified
Coding
Granite 4.2 8B: 47.7%Qwen3.5 397B: 76.2%Normalized gap 28.5Granite 4.2 8B source Qwen3.5 397B sourceGPQA
Knowledge
Granite 4.2 8B: 64.1%Qwen3.5 397B: 88.4%Normalized gap 24.3Granite 4.2 8B source Qwen3.5 397B sourceMMLU-Pro
Knowledge
Granite 4.2 8B: 74.0%Qwen3.5 397B: 87.8%Normalized gap 13.8Granite 4.2 8B source Qwen3.5 397B source
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
- Granite 4.2 8B
- Self-hosted; infrastructure cost varies
- Fits in one request
- Qwen3.5 397B
- $0.0024
- Fits in one request
Granite 4.2 8B has no comparable published API token rate.
Repository review
50K fresh input + 3K output tokens
- Granite 4.2 8B
- Self-hosted; infrastructure cost varies
- Fits in one request
- Qwen3.5 397B
- $0.0408
- Fits in one request
Granite 4.2 8B has no comparable published API token rate.
Cache-heavy agent loop
200K cached + 20K fresh input + 10K output tokens
- Granite 4.2 8B
- Self-hosted; infrastructure cost varies
- Does not fit 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
Granite 4.2 8B does not fit this 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. Granite 4.2 8B 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.
SpecificationGranite 4.2 8BQwen3.5 397B
Context window
Maximum documented context; output-token limits may be lower.
Granite 4.2 8B
Qwen3.5 397B
128K
API model ID
Granite 4.2 8B
ibm-granite/granite-4.2-8b
IBM Granite 4.2 8B 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.
Granite 4.2 8B
No comparable hosted API rate
IBM Granite 4.2 8B model cardQwen3.5 397B
Not published
Documented inputs
Granite 4.2 8B
Not sourced
Qwen3.5 397B
Not sourced
Documented outputs
Granite 4.2 8B
Not sourced
Qwen3.5 397B
Not sourced
Provider availability
Granite 4.2 8B
Not sourced
Qwen3.5 397B
Not sourced
Reasoning profile
Granite 4.2 8B
Reasoning
Qwen3.5 397B
Non-Reasoning
Weight access
Granite 4.2 8B
Open Weight
Qwen3.5 397B
Open Weight
License
Granite 4.2 8B
Open Weight
Qwen3.5 397B
Open Weight
Release date
Granite 4.2 8B
2026-08-25
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
- Qwen3.5 397B has the higher public score estimate, 58.2 versus 46.88, but the 90% score intervals overlap.
- Workload cost
- A complete comparable API-rate estimate is not available for both models.
- Context tradeoff
- Both models list 128K.
Run the same representative tasks against both endpoints before changing production traffic.
Benchmark evidence
The full public result ledger is available for audit without forcing a wide desktop table onto a phone.
Browse raw public benchmark evidence45 rows
Agentic
Terminal-Bench 2.1
Granite 4.2 8B20.6%SourceQwen3.5 397B—Not directly comparable
τ³-bench results
Granite 4.2 8B58.1%SourceQwen3.5 397B68.4%SourceQwen3.5 397B leads this result
BFCL v4
Granite 4.2 8B52.4%SourceQwen3.5 397B—Not directly comparable
Terminal-Bench 2.0
Granite 4.2 8B—Qwen3.5 397B52.5%SourceNot directly comparable
BrowseComp
Granite 4.2 8B—Qwen3.5 397B62%SourceNot directly comparable
Claw-Eval
Granite 4.2 8B—Qwen3.5 397B56.8%SourceNot directly comparable
QwenClawBench
Granite 4.2 8B—Qwen3.5 397B51.8%SourceNot directly comparable
VITA-Bench
Granite 4.2 8B—Qwen3.5 397B43.7%SourceNot directly comparable
DeepPlanning
Granite 4.2 8B—Qwen3.5 397B37.6%SourceNot directly comparable
Toolathlon
Granite 4.2 8B—Qwen3.5 397B36.3%SourceNot directly comparable
MCP Atlas
Granite 4.2 8B—Qwen3.5 397B46.1%SourceNot directly comparable
MCP-Tasks
Granite 4.2 8B—Qwen3.5 397B74.2%SourceNot directly comparable
WideResearch
Granite 4.2 8B—Qwen3.5 397B74.0%SourceNot directly comparable
Gert Labs
Granite 4.2 8B—Qwen3.5 397B46.76%SourceNot directly comparable
ResearchClawBench
Granite 4.2 8B—Qwen3.5 397B14.2%SourceNot directly comparable
Coding
SWE-bench Verified
Granite 4.2 8B47.7%SourceQwen3.5 397B76.2%SourceQwen3.5 397B leads this result
SWE-bench Pro
Granite 4.2 8B19.1%SourceQwen3.5 397B50.9%SourceQwen3.5 397B leads this result
SWE Multilingual
Granite 4.2 8B30.8%SourceQwen3.5 397B—Not directly comparable
Terminal-Bench 2.1
Granite 4.2 8B20.6%SourceQwen3.5 397B—Not directly comparable
LiveCodeBench v6
Granite 4.2 8B73.2%SourceQwen3.5 397B83.6%SourceQwen3.5 397B leads this result
SciCode
Granite 4.2 8B36.1%SourceQwen3.5 397B—Not directly comparable
Reasoning
Knowledge
GPQA
Granite 4.2 8B64.1%SourceQwen3.5 397B88.4%SourceQwen3.5 397B leads this result
MMLU-Pro
Granite 4.2 8B74.0%SourceQwen3.5 397B87.8%SourceQwen3.5 397B leads this result
SuperGPQA
Granite 4.2 8B—Qwen3.5 397B70.4%SourceNot directly comparable
MMLU-Redux
Granite 4.2 8B—Qwen3.5 397B94.9%SourceNot directly comparable
C-Eval
Granite 4.2 8B—Qwen3.5 397B93%SourceNot directly comparable
HLE
Granite 4.2 8B—Qwen3.5 397B28.7%SourceNot directly comparable
Math
AIME 2025
Granite 4.2 8B86.7%SourceQwen3.5 397B—Not directly comparable
HMMT Feb 2025
Granite 4.2 8B78.3%SourceQwen3.5 397B94.8%SourceQwen3.5 397B leads this result
AIME26
Granite 4.2 8B—Qwen3.5 397B93.3%SourceNot directly comparable
HMMT Nov 2025
Granite 4.2 8B—Qwen3.5 397B92.7%SourceNot directly comparable
HMMT Feb 2026
Granite 4.2 8B—Qwen3.5 397B87.9%SourceNot directly comparable
MMAnswerBench
Granite 4.2 8B—Qwen3.5 397B80.9%SourceNot directly comparable
Multilingual
Multimodal
MMMU-Pro
Granite 4.2 8B—Qwen3.5 397B79%SourceNot directly comparable
MathVision
Granite 4.2 8B—Qwen3.5 397B88.6%SourceNot directly comparable
CharXiv
Granite 4.2 8B—Qwen3.5 397B80.8%SourceNot directly comparable
VideoMMMU
Granite 4.2 8B—Qwen3.5 397B84.7%SourceNot directly comparable
ScreenSpot Pro
Granite 4.2 8B—Qwen3.5 397B65.6%SourceNot directly comparable
V*
Granite 4.2 8B—Qwen3.5 397B95.8%SourceNot directly comparable
Instruction following
Frequently asked questions
Which is better, Granite 4.2 8B or Qwen3.5 397B?
Qwen3.5 397B has the higher public score estimate, 58.2 versus 46.88, but the 90% score intervals overlap. The higher estimate is not a decisive winner because the uncertainty ranges overlap.
Which is better for coding, Granite 4.2 8B or Qwen3.5 397B?
The current coding averages use different weighted benchmark sets, so BenchLM does not name a winner from them. Read the shared benchmark rows directly and test the models on the same task set.
Which is better for agentic tasks, Granite 4.2 8B or Qwen3.5 397B?
The published evidence does not provide a shared weighted agentic tasks basis for both models, so BenchLM does not name a agentic tasks winner.
Which costs less, Granite 4.2 8B 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, Granite 4.2 8B or Qwen3.5 397B?
Both models list the same context window, 128K.
Related comparisons
Last updated August 31, 2026
Watch Granite 4.2 8B vs Qwen3.5 397B
One weekly email when material rank, price, or benchmark evidence changes make this matchup worth revisiting.
Read a sample issueJoin 2,000+ readers.