Across 13 comparable benchmarks, Qwen3.8-Max-0902 has better reported results on 9 and worse on 4. 3 further shared results are listed below, each with the reason it is not counted.
Moderate evidence13 of 16 shared results countedHow we compare
Price basis: Listed rates vs standard rates. The price section below shows the applicable periods.
| Fact | Qwen3.8-Max-0902 | Inkling 2 Small |
|---|---|---|
| Context window | 984K tokens | 1M tokens |
| Input / 1M | $2.00 | $0.300 |
| Output / 1M | $6.00 | $1.00 |
| Blended (4:1) | $1.54 | $0.246 |
| Status | Available | Available |
| Released | September 2, 2026 | November 5, 2026 |
Blended is 4:1 input to output with 90% cache hits at listed rates — set your own mix.
Mean of 8 shared benchmarks grouped by capability on a 0–100 scale. How averages are built
Qwen3.8-Max-0902 ahead on 3 of 5 counted
Qwen3.8-Max-0902 ahead on 1 of 1 counted
Qwen3.8-Max-0902 ahead on 2 of 3 counted
Split 1–1 of 2 counted
No fixed-scale average — see the rows in Benchmark results
No shared comparable results for Abstract reasoning, Mathematics, Computer & tool use, Research & search, Long context, Cybersecurity.
Published scores, with matching modes where available. Gaps show Qwen3.8-Max-0902 relative to Inkling 2 Small. Reading the results
| Benchmark | Qwen3.8-Max-0902 | Inkling 2 Small | Gap |
|---|---|---|---|
| DeepSWERepository Engineering · Pass@1 (DeepSWE v1.1) | 69.3Extra-High · tools | 72.1Extra-High · tools | −2.8 pts#27 vs #20 of 73 |
| SWE-MarathonRepository Engineering · Resolution rate (Pass@1) | 44.8Extra-High · tools | 38.8Extra-High · tools | +6.0 pts#7 vs #12 of 15 |
| Terminal-Bench 2.1Agentic Development · Accuracy | 88.8Thinking · tools | 85.1≠Extra-High · tools | +3.7 pts#11 vs #24 of 68 |
| Terminal-Bench 4.0Agentic Development · Resolution rate (%)Not counted · Different or unrecorded test environments — excluded from the tally | 38.9Unspecified Reasoning · tools· vals-mini-swe | 13.1≠Extra-High · tools· legacy-default | —#29 vs #57 of 90 |
| Terminal-Bench 4.1Agentic Development · Resolution rate (%) | 27.0Max · tools | 13.0≠Extra-High · tools | +13.9 pts#27 vs #38 of 48 |
| Vibe Code Bench v1.1Code Generation & Editing · Pass rate (%) | 77.1Unspecified Reasoning · tools | 81.6≠Extra-High · tools | −4.5 pts#40 vs #30 of 121 |
| BioMysteryBenchBiology & Genomics · Accuracy (%) | 84.1Max · tools | 70.0≠Extra-High · tools | +14.1 pts#2 vs #18 of 28 |
| BabyVisionVisual Understanding · Accuracy | 93.8Extra-High · tools | 93.7Extra-High · tools | +0.1 pts#6 vs #7 of 17 |
| ChartographyDocuments & Charts · Score | 42.0Max · tools | 80.4≠Extra-High · tools | 0.52×#25 vs #10 of 29 |
| MMMU-ProVisual Understanding · Accuracy | 82.8Thinking · no tools | 81.6≠Extra-High · no tools | +1.2 pts#16 vs #25 of 136 |
| JobBenchCross-industry Work · Mean task reward (%)Not counted · This metric has no better direction | 65.3Max · tools | 70.5≠Extra-High · tools | — |
| Legal Research BenchLegal · Score | 46.1Unspecified Reasoning · tools | 46.6≠Extra-High · tools | 0.99×#20 vs #18 of 91 |
| τ³-BankingService Workflows · Score | 47.8Thinking · tools | 40.4≠Extra-High · tools | 1.18×#9 vs #20 of 120 |
| HLEKnowledge Exams · AccuracyNot counted · Different or unrecorded test environments — excluded from the tally | 43.9Max · no tools· legacy-default | 44.1≠Extra-High · no tools · text only· text-only | —#70 vs #69 of 151 |
| AA Intelligence Index v5.0.1Capability Indices · Index (0-100) | 33.7Extra-High · tools | 30.6Extra-High · tools | 1.10×#24 vs #30 of 62 |
| Vals Index v2.2Capability Indices · GDP-weighted accuracy (%) | 59.0Reported best | 56.1≠Extra-High · tools | +2.9 pts#17 vs #21 of 36 |
Repository Engineering · Pass@1 (DeepSWE v1.1)
Repository Engineering · Resolution rate (Pass@1)
Agentic Development · Accuracy
Agentic Development · Resolution rate (%)
Not counted · Different or unrecorded test environments — excluded from the tally
Agentic Development · Resolution rate (%)
Code Generation & Editing · Pass rate (%)
Biology & Genomics · Accuracy (%)
Visual Understanding · Accuracy
Documents & Charts · Score
Visual Understanding · Accuracy
Cross-industry Work · Mean task reward (%)
Not counted · This metric has no better direction
Legal · Score
Service Workflows · Score
Knowledge Exams · Accuracy
Not counted · Different or unrecorded test environments — excluded from the tally
Capability Indices · Index (0-100)
Capability Indices · GDP-weighted accuracy (%)
A dash means no result is available.
| Benchmark | Qwen3.8-Max-0902 | Inkling 2 Small |
|---|---|---|
| FrontierMath Tier 4 v2 | 34.1 | — |
| FrontierMath v2 | 65.6 | — |
| GPQA Diamond | — | 93.0 |
| HLE-Diamond | — | 28.7 |
| HMMT Feb 2026 | — | 95.8 |
| MathArena Apex | — | 63.5 |
| Terminal-Bench-Science 0.1 | — | 5.7 |
| FrontierCS 2 | — | 16.4 |
| FrontierSWE v2 | 17.8 | — |
| MLS Bench | 50.1 | — |
| NL2Repo-Bench | 64.9 | — |
| ProgramBench | 28 | — |
| SWE-bench Multilingual | — | 81.2 |
| SWE-Bench Pro V2 | — | 97.7 |
| SWE-Bench Pro V2 Hard | — | 86.3 |
| SWE-bench Verified | — | 82.2 |
| Terminal-Bench 3.0 | 29.0 | — |
| Agents' Last Exam v1.1 | 27.9 | — |
| APEX-Agents | 63.3 | — |
| AutomationBench-AA | — | 45.6 |
| ClawEval-MM | 80.2 | — |
| Job Bench | 64.0 | — |
| MCP-Atlas | — | 85.4 |
| ScreenSpot-Pro (no tools) | — | 83.8 |
| Toolathlon-Verified | 73.3 | — |
| AA Intelligence Index v4.3 | 45.4 | — |
| Code Arena WebDev | 1676 | — |
| ECI | 155 | — |
| SuperCLUE | 72.6 | — |
| AA-LCR | — | 83.3 |
| LongBench v2 | — | 70.6 |
| CharXiv RQ | — | 87.7 |
| ERQA | 78.3 | — |
| GDP.pdf | — | 25.2 |
| MathVision | — | 92.8 |
| PerceptionBench | — | 64.7 |
| AA-Briefcase | — | 1527 |
| AutomationBench | 50.8 | — |
| CoWorkBench | 76.1 | — |
| EMB (Excel Modeling) | 72.0 | — |
| GDPval-AA v2 | — | 1630 |
| Tax Agent Bench | — | 65.8 |
| ARC-AGI-1 | — | 96.5 |
| ARC-AGI-2 | — | 81.7 |
| DeepSearchQA | — | 90.9 |
| AA-Omniscience | — | 16.1 |
Time-of-day providers use peak-hour rates; fixed-price providers use standard rates. Off-peak discounts remain in the published details.
Drag the handle above the bar for cache hits, the one below it for the input : output mix · 5% steps
Blended price = effective input rate at 90% cache hits × 80% + output rate × 20%
Qwen3.8-Max-0902 costs527%morethan Inkling 2 Small per million tokens — 6.3× the price.Using listed rates vs standard rates at this token mix.
Qwen3.8-Max-0902: International pricing. context_length <= 1000000 Alibaba Model Studio International
Inkling 2 Small: Standard pricing. Tinker 64K/256K inference; training billed separately. Image-specific token accounting not separately specified.
| Price / 1M tokens | Qwen3.8-Max-0902 | Inkling 2 Small | Ratio |
|---|---|---|---|
| Input | $2.00 | $0.30 | 6.67× |
| Output | $6.00 | $1.00 | 6.00× |
| Cached input | $0.25 | $0.03 | 8.33× |
| Cache write | $2.50 | — | — |
| Blended | $1.54 | $0.246 | 6.27× |
Estimates use the selected token mix. Cache writes, tool fees and long-context surcharges are excluded.
Reported cost per task, paired with the score from the same run. Settings and sources may differ; costs do not affect the verdict.
Qwen3.8-Max-0902 relative to Inkling 2 Small
Summary: 2 of 3 shared benchmarks have matching recorded settings and sources. The average is a geometric mean of 2 positive-cost ratios, one equal weight per benchmark. This benchmark set can include related tests and indexes; it is not a workload cost estimate.
Qwen3.8-Max-0902 8.5× more expensive
Qwen3.8-Max-0902 1.5× more expensive
Settings or sources differ
Qwen3.8-Max-0902 8.5× more expensive
Qwen3.8-Max-0902 1.5× more expensive
Settings or sources differ
Effort, protocol, aggregation and source match; other run conditions may differ.
Qwen3.8-Max-0902
Inkling 2 Small
Effort, protocol, aggregation and source match; other run conditions may differ.
Qwen3.8-Max-0902
Inkling 2 Small
Settings or sources differ or are not recorded; interpret the comparison with care.
Qwen3.8-Max-0902
Inkling 2 Small
| Benchmark | Qwen3.8-Max-0902 | Inkling 2 Small |
|---|---|---|
| AA-Briefcase | Not reported | |
| AA-Omniscience | Not reported | |
| ARC-AGI-1 | Not reported | |
| ARC-AGI-2 | Not reported | |
| AutomationBench-AA | Not reported | $0.0581 / task Average · Guardrail-safe objective completion rate: 45.6 Extra-High (With Tools) |
| EMB (Excel Modeling) | Not reported | |
| FrontierCS 2 | Not reported | |
| FrontierSWE v2 | Not reported | |
| GDP.pdf | Not reported | |
| HLE-Diamond | Not reported | |
| Legal Research Bench | Not reported | |
| Terminal-Bench-Science 0.1 | Not reported | |
| Vals Index v2.2 | $8.7295 / task Average · GDP-weighted accuracy (%): 59.0 Reported best (effort unspecified) | Not reported |
| Vibe Code Bench v1.1 | Not reported |
| Attribute | Qwen3.8-Max-0902 | Inkling 2 Small |
|---|---|---|
| Availability | Available | Available |
| Provider | Alibaba | Thinking Machines Lab |
| Model type | Reasoning model | Reasoning model |
| First release | September 2, 2026 | November 5, 2026 |
| Context window | 984K tokens | 1M tokens |
| Max output | 128K tokens | — |
| Total parameters | 2.4T | 594B |
| Active parameters | 95B | 17B |
| Architecture | Mixture of Experts (MoE) | Mixture of Experts (MoE) |
| Documented capabilities | Reasoning, Multilingual, Coding, Function calling, Tool use, Structured output, Web search, Long context | Reasoning, Coding, Tool use, Long context |
| Input modalities | Text, Image, Video | Text, Image |
| Output modalities | Text | Text |
| Weights | Closed / API only | Open weights |
| Licence | — | Apache 2.0 |
| Commercial use | Proprietary | Commercial use allowed |
These are reported-score comparisons, not controlled head-to-head tests.
What people found comparing these models, or using either one. Comments also appear on the models' own pages.
Used Qwen3.8-Max-0902 or Inkling 2 Small? Sign in with Google to share what worked, what did not, and which API or setup you tested.
No community comments on either model yet.
Alternatives with the most benchmark overlap for this matchup.