Model comparison
Gemini 3.5 Flash vs Sakana Fugu-Ultra
Head-to-head evidence from 8 shared benchmark results across 5 categories. Overall scores shown here use the public BenchAlign v5 ranking lane.
Public leaderboard positions: Gemini 3.5 Flash #33 (Estimated); Sakana Fugu-Ultra unranked (Not scored). Intervals and evidence labels describe ranking uncertainty, not a guarantee for a specific workload.
Evidence parity. Gemini 3.5 Flash and Sakana Fugu-Ultra share 8 comparable benchmark results. 5 of 8 categories are comparable. 38 results are unique to Gemini 3.5 Flash; 3 to Sakana Fugu-Ultra.
Updated July 23, 2026- Shared results
- 8
- Gemini 3.5 Flash only
- 38
- Sakana Fugu-Ultra only
- 3
- Comparable categories
- 5 / 8
Treat this as a split decision. Gemini 3.5 Flash makes more sense if its workflow fits your team better; Sakana Fugu-Ultra is the better fit if knowledge is the priority.
Confidence note. This is a partial-evidence comparison with 8 shared benchmark results across 5 evidence categories; 5 of 8 categories currently have scoreable aggregates for both models. Treat the verdict as directional until coverage is more balanced.
Why this result
Gemini 3.5 Flash and Sakana Fugu-Ultra finish on the same BenchAlign overall score, so this is less about a single winner and more about where the edge shows up. The BenchAlign headline says tie; the benchmark table is where the real choice happens.
Category breakdown
Exact category averages are shown below. Not measured means BenchLM does not have enough sourced public coverage for that model and category.
| Category | Gemini 3.5 Flash | Δ | Sakana Fugu-Ultra |
|---|---|---|---|
| Knowledge | Gemini 3.5 Flash47.2 | Margin→ 48.3 | Sakana Fugu-Ultra95.5 |
| Reasoning | Gemini 3.5 Flash74.7 | Margin→ 18.9 | Sakana Fugu-Ultra93.6 |
| Coding | Gemini 3.5 Flash53.9 | Margin→ 10.6 | Sakana Fugu-Ultra64.5 |
| Agentic | Gemini 3.5 Flash77.2 | Margin→ 4.9 | Sakana Fugu-Ultra82.1 |
| Multimodal | Gemini 3.5 Flash83.8 | Margin→ 2.8 | Sakana Fugu-Ultra86.6 |
| Math | Gemini 3.5 Flash32.9 | MarginNo overlap | Sakana Fugu-UltraNot measured |
| Inst. Following | Gemini 3.5 Flash76.3 | MarginNo overlap | Sakana Fugu-UltraNot measured |
Decisive benchmark drivers
The largest measured benchmark gaps in this matchup, with exact reported values.
More
- Source ↗
SWE-bench Pro
CodingA 55.1%B 73.7%Winner: Sakana Fugu-UltraΔ 18.6SWE-bench Pro: Gemini 3.5 Flash scored 55.1%; Sakana Fugu-Ultra scored 73.7%. Sakana Fugu-Ultra wins this benchmark. - Source ↗
MRCRv2
ReasoningA 77.3%B 93.6%Winner: Sakana Fugu-UltraΔ 16.3MRCRv2: Gemini 3.5 Flash scored 77.3%; Sakana Fugu-Ultra scored 93.6%. Sakana Fugu-Ultra wins this benchmark. - Source ↗
Terminal-Bench 2.0
AgenticA 76.2%B 82.1%Winner: Sakana Fugu-UltraΔ 5.9Terminal-Bench 2.0: Gemini 3.5 Flash scored 76.2%; Sakana Fugu-Ultra scored 82.1%. Sakana Fugu-Ultra wins this benchmark. - Source ↗
SciCode
CodingA 53.1%B 58.7%Winner: Sakana Fugu-UltraΔ 5.6SciCode: Gemini 3.5 Flash scored 53.1%; Sakana Fugu-Ultra scored 58.7%. Sakana Fugu-Ultra wins this benchmark. - Source ↗
GPQA
KnowledgeA 92.2%B 95.5%Winner: Sakana Fugu-UltraΔ 3.3GPQA: Gemini 3.5 Flash scored 92.2%; Sakana Fugu-Ultra scored 95.5%. Sakana Fugu-Ultra wins this benchmark.
Operational comparison
Runtime and commercial metrics are compared only when both models have a complete sourced value.
| Metric | Gemini 3.5 Flash | Sakana Fugu-Ultra | Comparison |
|---|---|---|---|
| Input / output priceUSD per 1M tokens | Gemini 3.5 Flash$1.5 input / $9 output | Sakana Fugu-UltraNot available | A complete price comparison is not available. |
| Generation speedtokens per second | Gemini 3.5 Flash284.2 tok/s | Sakana Fugu-UltraNot available | A complete speed comparison is not available. |
| First-answer latencyseconds to first token | Gemini 3.5 Flash18.55 s | Sakana Fugu-UltraNot available | A complete latency comparison is not available. |
| Context windowmaximum listed tokens | Gemini 3.5 Flash1M | Sakana Fugu-Ultra1M | Listed context windows are equal. |
Benchmark Deep Dive
AgenticSakana Fugu-Ultra wins15 benchmarks
| Benchmark | Gemini 3.5 Flash | Sakana Fugu-Ultra | Result |
|---|---|---|---|
| Terminal-Bench 2.0Source | 76.2% | 82.1% | Sakana Fugu-Ultra leads |
| MCP AtlasSource | 83.6% | — | Not comparable |
| ToolathlonSource | 56.5% | — | Not comparable |
| OSWorld-VerifiedSource | 78.4% | — | Not comparable |
| Finance Agent v2Source | 57.9% | — | Not comparable |
| GDPval-AASource | 1349 | — | Not comparable |
| τ²-bench resultsSource | 95.3% | — | Not comparable |
| GDPval-AASource | 42.4% | — | Not comparable |
| AA Agentic IndexSource | 37.5% | — | Not comparable |
| APEX-Agents-AASource | 47.1% | — | Not comparable |
| Gert LabsSource | 61.85% | — | Not comparable |
| ResearchClawBenchSource | 18.0% | — | Not comparable |
| AA AutomationBenchSource | 42.6% | — | Not comparable |
| AA EnterpriseOps-GymSource | 50.1% | — | Not comparable |
| terminalBenchHardSource | 40.9% | — | Not comparable |
CodingSakana Fugu-Ultra wins10 benchmarks
| Benchmark | Gemini 3.5 Flash | Sakana Fugu-Ultra | Result |
|---|---|---|---|
| Terminal-Bench 2.0Source | 76.2% | 82.1% | Sakana Fugu-Ultra leads |
| SWE-bench ProSource | 55.1% | 73.7% | Sakana Fugu-Ultra leads |
| SciCodeSource | 53.1% | 58.7% | Sakana Fugu-Ultra leads |
| Vibe Code BenchSource | 48.68% | — | Not comparable |
| cursorBench31Source | 49.8% | — | Not comparable |
| cursorBench32Source | 48.8% | — | Not comparable |
| AA Coding IndexSource | 70.1% | — | Not comparable |
| AA-SciCodeSource | 53.1% | — | Not comparable |
| LiveCodeBench v6Source | — | 93.2% | Not comparable |
| LiveCodeBench ProSource | — | 90.8% | Not comparable |
ReasoningSakana Fugu-Ultra wins5 benchmarks
KnowledgeSakana Fugu-Ultra wins10 benchmarks
| Benchmark | Gemini 3.5 Flash | Sakana Fugu-Ultra | Result |
|---|---|---|---|
| Artificial Analysis Intelligence IndexSource | 50.2% | — | Not comparable |
| GPQASource | 92.2% | 95.5% | Sakana Fugu-Ultra leads |
| GPQA-DSource | 92.7% | 95.5% | Sakana Fugu-Ultra leads |
| HLESource | 40.2% | — | Not comparable |
| AA-Omniscience AccuracySource | 51.9% | — | Not comparable |
| AA-Omniscience Hallucination RateSource | 60.7% | — | Not comparable |
| AA-GPQA DiamondSource | 92.2% | — | Not comparable |
| AA-HLESource | 41.0% | — | Not comparable |
| AA-Omniscience IndexSource | 22.7% | — | Not comparable |
| HLE w/o toolsSource | — | 50% | Not comparable |
Math2 benchmarks
MultimodalSakana Fugu-Ultra wins5 benchmarks
Frequently Asked Questions (6)
Which is better, Gemini 3.5 Flash or Sakana Fugu-Ultra?
Gemini 3.5 Flash and Sakana Fugu-Ultra are tied on the BenchAlign overall score, so the right pick depends on which category matters most for your use case.
Which is better for knowledge tasks, Gemini 3.5 Flash or Sakana Fugu-Ultra?
Sakana Fugu-Ultra has the edge for knowledge tasks in this comparison, averaging 95.5 versus 47.2. Inside this category, GPQA is the benchmark that creates the most daylight between them.
Which is better for coding, Gemini 3.5 Flash or Sakana Fugu-Ultra?
Sakana Fugu-Ultra has the edge for coding in this comparison, averaging 64.5 versus 53.9. Inside this category, SWE-bench Pro is the benchmark that creates the most daylight between them.
Which is better for reasoning, Gemini 3.5 Flash or Sakana Fugu-Ultra?
Sakana Fugu-Ultra has the edge for reasoning in this comparison, averaging 93.6 versus 74.7. Inside this category, MRCRv2 is the benchmark that creates the most daylight between them.
Which is better for agentic tasks, Gemini 3.5 Flash or Sakana Fugu-Ultra?
Sakana Fugu-Ultra has the edge for agentic tasks in this comparison, averaging 82.1 versus 77.2. Inside this category, Terminal-Bench 2.0 is the benchmark that creates the most daylight between them.
Which is better for multimodal and grounded tasks, Gemini 3.5 Flash or Sakana Fugu-Ultra?
Sakana Fugu-Ultra has the edge for multimodal and grounded tasks in this comparison, averaging 86.6 versus 83.8. Inside this category, CharXiv is the benchmark that creates the most daylight between them.
Related Comparisons
Explore More
Choose a model with this week’s evidence
Join 2,000+ readers for ranking moves, pricing changes, and the claims that still need proof.
One email each week. Unsubscribe anytime.