# Qwen3.8-27B vs Sakana Fugu-Ultra

> Side-by-side benchmark comparison for Qwen3.8-27B and Sakana Fugu-Ultra across agentic, coding, multimodal, knowledge, reasoning, multilingual, and math tasks.

- Canonical page: https://benchlm.ai/compare/qwen3-8-27b-vs-sakana-fugu-ultra
- Last updated: September 28, 2026

- Shared sourced benchmarks: 8
- HTML indexing: indexable
- Ranking lane: BenchAlign v5.7

## Quick Verdict

Pick Sakana Fugu-Ultra if you want the stronger benchmark profile. Qwen3.8-27B only makes more sense when its price, context window, or workload-specific category wins matter more than the overall score.

## Summary

- Sakana Fugu-Ultra leads overall 74 to 55.26.
- The clearest category separation is in multimodal & grounded, where the averages are 80.9 for Qwen3.8-27B and 73.6 for Sakana Fugu-Ultra.
- The biggest single benchmark swing is HLE w/o tools in Knowledge, with scores of 30.8% and 50%.
- Sakana Fugu-Ultra is the cheaper option on output tokens, which matters if you expect large responses or heavy interactive use.
- Sakana Fugu-Ultra also has the larger context window at 1M.

## Model Snapshot

| Property | Qwen3.8-27B | Sakana Fugu-Ultra |
|----------|----------|----------|
| Creator | Alibaba | Sakana AI |
| Type | Open Weight | Proprietary |
| Reasoning | Reasoning | Reasoning |
| Context | 262K | 1M |
| Overall Score | 55.26 | null |
| Benchmarks Covered | 33 | 11 |
| Pricing (input/output) | $0.00 / $0.00 | Pricing unavailable / Pricing unavailable |

## Category Breakdown

### Agentic

- Winner: Not comparable
- Qwen3.8-27B public-lane score: 61.2 (Supported · #18/117)
- Sakana Fugu-Ultra public-lane score: Coming soon

| Benchmark | Qwen3.8-27B | Sakana Fugu-Ultra | Winner |
|-----------|-----------|-----------|--------|
| Terminal-Bench 2.1 | 73.0% | 82.1% | Sakana Fugu-Ultra |
| CoWorkBench | 70.7% | Coming soon | Coming soon |
| JobBench | 33.4% | Coming soon | Coming soon |
| Agents' Last Exam | 42.9% | Coming soon | Coming soon |
| OSWorld-Verified | 84.3% | Coming soon | Coming soon |
| WebArena-Verified | 64.8% | Coming soon | Coming soon |
| AndroidWorld | 81.9% | Coming soon | Coming soon |
| Terminal-Bench 2.1 (Vals) | 58.4% | Coming soon | Coming soon |

### Coding

- Winner: Not comparable
- Qwen3.8-27B public-lane score: 48.7 (Supported · #45/142)
- Sakana Fugu-Ultra public-lane score: Coming soon

| Benchmark | Qwen3.8-27B | Sakana Fugu-Ultra | Winner |
|-----------|-----------|-----------|--------|
| Terminal-Bench 2.1 | 73.0% | 82.1% | Sakana Fugu-Ultra |
| SWE-bench Pro | 61.7% | 73.7% | Sakana Fugu-Ultra |
| NL2Repo | 42.3% | Coming soon | Coming soon |
| DeepSWE | 42.2% | Coming soon | Coming soon |
| LiveCodeBench v6 | 90.3% | 93.2% | Sakana Fugu-Ultra |
| VulcanBench v3 | 82.6% | Coming soon | Coming soon |
| LiveCodeBench (Vals) | 84.0% | Coming soon | Coming soon |
| SWE-bench (Vals) | 86.0% | Coming soon | Coming soon |
| LiveCodeBench Pro | Coming soon | 90.8% | Coming soon |
| SciCode | Coming soon | 58.7% | Coming soon |

### Reasoning

- Winner: Not comparable
- Qwen3.8-27B public-lane score: 78.7 (#8/27)
- Sakana Fugu-Ultra public-lane score: 79.1 (Unranked · 1 rankable row)

| Benchmark | Qwen3.8-27B | Sakana Fugu-Ultra | Winner |
|-----------|-----------|-----------|--------|
| MRCRv2 | Coming soon | 93.6% | Coming soon |

### Multimodal & Grounded

- Winner: Not comparable
- Qwen3.8-27B public-lane score: 80.9 (#11/50)
- Sakana Fugu-Ultra public-lane score: 73.6 (Unranked · 1 rankable row)

| Benchmark | Qwen3.8-27B | Sakana Fugu-Ultra | Winner |
|-----------|-----------|-----------|--------|
| MathVision | 90.0% | Coming soon | Coming soon |
| MathVision w/ Python | 94.6% | Coming soon | Coming soon |
| BabyVision | 65.7% | Coming soon | Coming soon |
| BabyVision w/ Python | 85.6% | Coming soon | Coming soon |
| Vision2Web | 62.9% | Coming soon | Coming soon |
| CharXiv w/o tools | 83.7% | Coming soon | Coming soon |
| CharXiv | 90.2% | 86.6% | Qwen3.8-27B |
| OmniDocBench 1.5 | 91.1% | Coming soon | Coming soon |
| RealWorldQA | 85.9% | Coming soon | Coming soon |
| ERQA | 65.5% | Coming soon | Coming soon |

### Knowledge

- Winner: Not comparable
- Qwen3.8-27B public-lane score: 49.2 (Supported · #63/168)
- Sakana Fugu-Ultra public-lane score: Coming soon

| Benchmark | Qwen3.8-27B | Sakana Fugu-Ultra | Winner |
|-----------|-----------|-----------|--------|
| GPQA | 89.2% | 95.5% | Sakana Fugu-Ultra |
| GPQA-D | 89.2% | 95.5% | Sakana Fugu-Ultra |
| HLE | 30.8% | Coming soon | Coming soon |
| HLE w/o tools | 30.8% | 50% | Sakana Fugu-Ultra |
| GPQA Diamond (Vals) | 88.9% | Coming soon | Coming soon |
| MMLU-Pro (Vals) | 84.3% | Coming soon | Coming soon |

### Instruction Following

- Winner: Not comparable
- Qwen3.8-27B public-lane score: 83.2 (#45/124)
- Sakana Fugu-Ultra public-lane score: Coming soon

| Benchmark | Qwen3.8-27B | Sakana Fugu-Ultra | Winner |
|-----------|-----------|-----------|--------|
| IFBench | 79.5% | Coming soon | Coming soon |

## FAQ

### Which is better overall, Qwen3.8-27B or Sakana Fugu-Ultra?

Sakana Fugu-Ultra is ahead overall on BenchLM right now.

### Where is the biggest gap between Qwen3.8-27B and Sakana Fugu-Ultra?

The widest category gap is in multimodal & grounded, where the averages are 80.9 for Qwen3.8-27B and 73.6 for Sakana Fugu-Ultra.

### How many benchmarks does Qwen3.8-27B cover on BenchLM?

Qwen3.8-27B currently has 33 sourced benchmark scores on BenchLM.

### How many benchmarks does Sakana Fugu-Ultra cover on BenchLM?

Sakana Fugu-Ultra currently has 11 sourced benchmark scores on BenchLM.

## Related Comparisons

- [Qwen3.8-27B vs Sakana Fugu](/compare/qwen3-8-27b-vs-sakana-fugu)
- [Sakana Fugu-Ultra vs Sakana Fugu](/compare/sakana-fugu-vs-sakana-fugu-ultra)
- [Qwen3.8-27B vs Fugu Cyber](/compare/qwen3-8-27b-vs-sakana-fugu-cyber)
- [Sakana Fugu-Ultra vs Fugu Cyber](/compare/sakana-fugu-cyber-vs-sakana-fugu-ultra)

## Explore More

- [Qwen3.8-27B profile](/models/qwen3-8-27b)
- [Sakana Fugu-Ultra profile](/models/sakana-fugu-ultra)
- [Compare Pricing](/llm-pricing)
- [Alternative Finder](/tools/alternative-finder)
- [LLM Selector](/tools/llm-selector)
- [Overall Rankings](/best/overall)