Across 21 comparable benchmarks, Inkling 2 Mini has better reported results on 14 and worse on 7. 2 further shared results are listed below, each with the reason it is not counted.
Strong evidence21 of 23 shared results countedHow we compare
Price basis: Listed rates vs standard rates. The price section below shows the applicable periods.
| Fact | Qwen3.8-27B | Inkling 2 Mini |
|---|---|---|
| Context window | 256K tokens | 1M tokens |
| Input / 1M | $0.500 | $0.150 |
| Output / 1M | $3.00 | $0.500 |
| Blended (4:1) | $0.712 | $0.123 |
| Status | Available | Available |
| Released | August 14, 2026 | November 20, 2026 |
Blended is 4:1 input to output with 90% cache hits at listed rates — set your own mix.
Mean of 14 shared benchmarks grouped by capability on a 0–100 scale. How averages are built
Inkling 2 Mini ahead on 3 of 3 counted
Qwen3.8-27B ahead on 1 of 1 counted
Inkling 2 Mini ahead on 2 of 3 counted
Inkling 2 Mini ahead on 2 of 2 counted
Inkling 2 Mini ahead on 2 of 2 counted
Inkling 2 Mini ahead on 5 of 6 counted
Qwen3.8-27B ahead on 2 of 2 counted
No fixed-scale average — see the rows in Benchmark results
No shared comparable results for Mathematics, Science & health, Cybersecurity.
Published scores, with matching modes where available. Gaps show Qwen3.8-27B relative to Inkling 2 Mini. Reading the results
| Benchmark | Qwen3.8-27B | Inkling 2 Mini | Gap |
|---|---|---|---|
| DeepSWERepository Engineering · Pass@1 (DeepSWE v1.1) | 42.2Thinking · tools | 48.2Thinking · tools | −6.0 pts#64 vs #61 of 73 |
| FrontierCode 2.0Repository Engineering · Weighted rubric score | 14.9Thinking · tools | 15.1Thinking · tools | 0.99×#27 vs #26 of 36 |
| Vibe Code Bench v1.1Code Generation & Editing · Pass rate (%) | 64.8Extra-High · tools | 77.9≠Thinking · tools | −13.1 pts#54 vs #38 of 121 |
| ARC-AGI-1Abstract Generalization · Score (%) | 87.5Extra-High · no tools | 84.5≠Thinking · no tools | +3.0 pts#40 vs #47 of 87 |
| ARC-AGI-2Abstract Generalization · Score (% solved); cost per task (USD)Not counted · Speed and cost telemetry is reported, not scored | 42.4Extra-High · no tools | 49.3≠Thinking · no tools | 0.86×#48 vs #45 of 80 |
| AutomationBench-AATool Orchestration · Guardrail-safe objective completion rate | 48.2Thinking · tools | 41.8Thinking · tools | +6.4 pts#15 vs #17 of 18 |
| MCP-AtlasTool Orchestration · Pass rate / claim coverage | 77.3Thinking · tools | 82.7Thinking · tools | −5.4 pts#25 vs #14 of 55 |
| ScreenSpot-Pro (no tools)GUI Grounding · Success rate (%) | 57.6Thinking · no tools | 76.5Thinking · no tools | −18.9 pts#6 vs #4 of 8 |
| BrowseCompFact Finding · Accuracy | 61.6Thinking · tools | 79.0Thinking · tools | −17.4 pts#40 vs #24 of 61 |
| DeepSearchQAFact Finding · F1 score | 79.7Thinking · tools | 85.3Thinking · tools | 0.93×#6 vs #4 of 9 |
| LongBench v2Long Reasoning · Accuracy | 52.3Thinking · no tools | 66.9Thinking · no tools | −14.6 pts#17 vs #2 of 22 |
| OpenAI MRCR v2 (8-needle, 256K–512K)Long Retrieval · Accuracy (%) | 38.0Thinking · no tools | 46.4Thinking · no tools | −8.4 pts#4 vs #3 of 5 |
| ChartographyDocuments & Charts · Score | 52.1Thinking · tools | 72.5Thinking · tools | 0.72×#24 vs #16 of 29 |
| GDP.pdfDocuments & Charts · Task pass rate (%) | 16.4Extra-High · no tools | 16.8≠Thinking · no tools | −0.4 pts#39 vs #38 of 96 |
| MMMU-ProVisual Understanding · Accuracy | 76.3Extra-High · no tools | 74.9≠Thinking · no tools | +1.4 pts#58 vs #69 of 136 |
| OmniDocBenchDocuments & Charts · Accuracy | 91.1Thinking · no tools | 92.8Thinking · no tools | −1.7 pts#2 vs #1 of 6 |
| PerceptionBenchVisual Understanding · Accuracy | 52.4Thinking · no tools | 57.5Thinking · no tools | −5.1 pts#14 vs #11 of 19 |
| ZeroBench MainVisual Understanding · Pass@1 / Pass@5 / Pass^5 | 12.0Thinking · no tools | 34.0≠Thinking · tools | −22.0 pts#19 vs #16 of 21 |
| AA-BriefcaseOffice & Business · AA-Briefcase Elo | 1391Thinking · tools | 1263Thinking · tools | 1.10×#17 vs #18 of 20 |
| GDPval-AA v2Cross-industry Work · Elo score | 1427Thinking · tools | 1319Thinking · tools | 1.08×#17 vs #19 of 23 |
| JobBenchCross-industry Work · Mean task reward (%)Not counted · This metric has no better direction | 33.4Thinking · tools | 49.6Thinking · tools | — |
| HLE-DiamondKnowledge Exams · Accuracy (%) | 15.3Thinking · no tools | 15.2Thinking · no tools | +0.1 pts#12 vs #13 of 16 |
| AA Intelligence Index v5.0.1Capability Indices · Index (0-100) | 20.1Extra-High · tools | 20.1≠Thinking · no tools | 1.00×#43 vs #44 of 62 |
Repository Engineering · Pass@1 (DeepSWE v1.1)
Repository Engineering · Weighted rubric score
Code Generation & Editing · Pass rate (%)
Tool Orchestration · Guardrail-safe objective completion rate
Tool Orchestration · Pass rate / claim coverage
GUI Grounding · Success rate (%)
Fact Finding · Accuracy
Fact Finding · F1 score
Long Reasoning · Accuracy
Long Retrieval · Accuracy (%)
Documents & Charts · Score
Documents & Charts · Task pass rate (%)
Visual Understanding · Accuracy
Documents & Charts · Accuracy
Visual Understanding · Accuracy
Visual Understanding · Pass@1 / Pass@5 / Pass^5
Office & Business · AA-Briefcase Elo
Cross-industry Work · Elo score
Cross-industry Work · Mean task reward (%)
Not counted · This metric has no better direction
Knowledge Exams · Accuracy (%)
Capability Indices · Index (0-100)
A dash means no result is available.
| Benchmark | Qwen3.8-27B | Inkling 2 Mini |
|---|---|---|
| CritPt | 5.4 | — |
| GPQA Diamond | 89.2 | — |
| HLE | 30.8 | — |
| ProofBench v1.1 (Lean 4) | 16.0 | — |
| SciCode | 46.6 | — |
| AA Cyber Index | 12.7 | — |
| APEX-SWE | 50.7 | — |
| Code Migration | 14.2 | — |
| IOI (Vals v2) | 39.1 | — |
| LiveCodeBench | 90.3 | — |
| NL2Repo-Bench | 42.3 | — |
| SWE-Bench Pro - Public | 61.7 | — |
| Terminal-Bench 2.1 | 73.0 | — |
| Terminal-Bench 4.0 | 5.6 | — |
| Agents' Last Exam v1.0 | 20.4 | — |
| APEX-Agents | 47.5 | — |
| OSWorld-Verified | 84.3 | — |
| ECI | 149 | — |
| LiveBench | 75.3 | — |
| LiveBench 2.0 | 62.7 | — |
| Vals Index | 48.5 | — |
| Vals Index v2.2 | — | 46.5 |
| Creative Writing | 1671 | — |
| Context Arena | 94.0 | — |
| BabyVision | 85.6 | — |
| CharXiv RQ | 90.2 | — |
| GeoBench ACW | 38.0 | — |
| MathVision | 94.6 | — |
| AA-AnalystAgent | — | 46.3 |
| EMB (Excel Modeling) | 59.7 | — |
| Finance Agent v2 | 48.5 | — |
| Harvey's Legal Agent Benchmark | 11.3 | — |
| Legal Research Bench | 36.1 | — |
| SAGE | 52.4 | — |
| Tax Agent Bench | 30.9 | — |
| τ³-Banking | 48.0 | — |
| SimpleBench | 60.2 | — |
| MedCode | 28.7 | — |
| MedScribe | 83.8 | — |
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%
Qwen3.8-27B costs480%morethan Inkling 2 Mini per million tokens — 5.8× the price.Using listed rates vs standard rates at this token mix.
Qwen3.8-27B: International pricing. context_length <= 1000000 Alibaba Model Studio International (Singapore)
Inkling 2 Mini: Standard pricing. Tinker 64K/256K training (1M sampling); fine-tuning billed separately ($0.75/1M for 64K, $1.50/1M for 256K).
| Price / 1M tokens | Qwen3.8-27B | Inkling 2 Mini | Ratio |
|---|---|---|---|
| Input | $0.50 | $0.15 | 3.33× |
| Output | $3.00 | $0.50 | 6.00× |
| Cached input | $0.10 | $0.015 | 6.67× |
| Cache write | — | — | — |
| Blended | $0.712 | $0.123 | 5.80× |
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.
| Benchmark | Qwen3.8-27B | Inkling 2 Mini |
|---|---|---|
| AA Intelligence Index v5.0.1 | 1 more settings | Not reported |
| ARC-AGI-1 | 1 more settings | Not reported |
| ARC-AGI-2 | 1 more settings | Not reported |
| LiveBench 2.0 | $0.519 / task Average · Overall score (equal-weight mean of 7 suite scores, 0-100): 62.7 Max (task-specific tools) | Not reported |
| Tax Agent Bench | $0.8153 / task Average · Accuracy (%, rubric × citation quality): 30.9 Extra-High (With Tools) | Not reported |
| Attribute | Qwen3.8-27B | Inkling 2 Mini |
|---|---|---|
| Availability | Available | Available |
| Provider | Alibaba | Thinking Machines Lab |
| Model type | Reasoning model | Reasoning model |
| First release | August 14, 2026 | November 20, 2026 |
| Context window | 256K tokens | 1M tokens |
| Max output | 128K tokens | — |
| Total parameters | 27B | 26.4B |
| Active parameters | 27B | 5.4B |
| Architecture | Dense | Mixture of Experts (MoE) |
| Documented capabilities | Reasoning, Multilingual, Coding, Tool use, Function calling, Long context, Computer use | Reasoning, Coding, Tool use, Long context |
| Input modalities | Text, Image, Video | Text, Image |
| Output modalities | Text | Text |
| Weights | Open weights | Open weights |
| Licence | Apache 2.0 | Apache 2.0 |
| Commercial use | Commercial use allowed | 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 Qwen3.8-27B or Inkling 2 Mini? 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.