Model comparison
GPT-5.4 mini vs Tulu 3 (Tülu 3) 70B
GPT-5.4 mini is the stronger model overall, scoring 44.8 to 32.6 on the Noometry Index.
Last verified . 13 shared benchmarks.
Allen Institute for AI (Ai2)Summary
- They share 13 benchmarks with published results for both. GPT-5.4 mini scores higher in 8 categories and Tulu 3 (Tülu 3) 70B in 0 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in math, where GPT-5.4 mini leads 45.4 to 13.3.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 88.9% for GPT-5.4 mini and 4.4% for Tulu 3 (Tülu 3) 70B.
- Tulu 3 (Tülu 3) 70B has downloadable open weights; the other is API-only.
Side by side
| GPT-5.4 mini | Tulu 3 (Tülu 3) 70B | |
|---|---|---|
| Provider | OpenAI | Allen Institute for AI (Ai2) |
| Noometry Index | 44.8 | 32.6 |
| Released | 2026-03-17 | 2024-11-21 |
| Weights | Proprietary | Open |
| Context window | 400K | — |
| Max output | 128K | — |
| Input $ / M tokens | $0.75 | — |
| Output $ / M tokens | $4.50 | — |
| Results tracked | 46 | 14 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5.4 mini leads
GPT-5.4 mini: 45.0 (#67), Tulu 3 (Tülu 3) 70B: 35.4 (#211)
| Benchmark | GPT-5.4 mini | Tulu 3 (Tülu 3) 70B |
|---|---|---|
| LMArena Coding | 1438 | 1236 |
| FrontierCode | 27% | — |
| LMArena WebDev | 1397 | — |
| SciCode | 49.9% | — |
| WeirdML | 60.3% | — |
| ALE-Bench | 1,189 | — |
Agentic & Tool Use Not comparable
GPT-5.4 mini: 29.6 (#74), Tulu 3 (Tülu 3) 70B: —
| Benchmark | GPT-5.4 mini | Tulu 3 (Tülu 3) 70B |
|---|---|---|
| DeepResearch Bench | 36.3% | — |
Reasoning GPT-5.4 mini leads
GPT-5.4 mini: 30.3 (#77), Tulu 3 (Tülu 3) 70B: 22.9 (#172)
| Benchmark | GPT-5.4 mini | Tulu 3 (Tülu 3) 70B |
|---|---|---|
| LMArena Hard Prompts | 1424 | 1220 |
| ARC-AGI-2 | 18.9% | — |
| Kagi LLM Benchmark | 37.9% | — |
| NYT Connections (extended) | 61.8% | — |
| ARC-AGI-1 | 63.7% | — |
| CritPt | 10% | — |
| Chess Puzzles | 24% | — |
| Thematic Generalization | 61.7% | — |
| Mystery Game Puzzles | 11% | — |
| DTBench | 80% | — |
| LMCA | 40.8% | — |
| Epoch Capabilities Index | 148.83 | — |
| ForecastBench | 57 | — |
Math GPT-5.4 mini leads
GPT-5.4 mini: 45.4 (#63), Tulu 3 (Tülu 3) 70B: 13.3 (#297)
| Benchmark | GPT-5.4 mini | Tulu 3 (Tülu 3) 70B |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 88.9% | 4.4% |
| LMArena Math | 1418 | 1242 |
| FrontierMath (Tiers 1-3) | 51.2% | — |
| FrontierMath Tier 4 | 9.8% | — |
| ProofBench | 21% | — |
| MATH Level 5 | — | 42.7% |
| FrontierMath (Feb 2025 set) | 28.3% | — |
| FrontierMath Tier 4 (v1) | 2.1% | — |
Knowledge GPT-5.4 mini leads
GPT-5.4 mini: 51.4 (#62), Tulu 3 (Tülu 3) 70B: 25.1 (#257)
| Benchmark | GPT-5.4 mini | Tulu 3 (Tülu 3) 70B |
|---|---|---|
| GPQA Diamond | 86.9% | 46.3% |
| SimpleQA Verified | 29.4% | — |
| Vectara Hallucination Rate | 5.5% | — |
| LMArena Expert | 1434 | — |
Multimodal Not comparable
GPT-5.4 mini: 39.0 (#51), Tulu 3 (Tülu 3) 70B: —
| Benchmark | GPT-5.4 mini | Tulu 3 (Tülu 3) 70B |
|---|---|---|
| LMArena Vision | 1245 | — |
Multilingual GPT-5.4 mini leads
GPT-5.4 mini: 52.1 (#94), Tulu 3 (Tülu 3) 70B: 40.1 (#217)
| Benchmark | GPT-5.4 mini | Tulu 3 (Tülu 3) 70B |
|---|---|---|
| LMArena Non-English | 1405 | 1236 |
| LMArena Chinese | 1446 | 1249 |
| LMArena Russian | 1417 | 1246 |
| LMArena French | 1440 | — |
| LMArena German | 1407 | — |
| LMArena Japanese | 1375 | — |
| LMArena Korean | 1369 | — |
| LMArena Spanish | 1403 | — |
Instruction Following GPT-5.4 mini leads
GPT-5.4 mini: 74.3 (#99), Tulu 3 (Tülu 3) 70B: 65.0 (#222)
| Benchmark | GPT-5.4 mini | Tulu 3 (Tülu 3) 70B |
|---|---|---|
| LMArena Instruction Following | 1405 | 1233 |
Long Context GPT-5.4 mini leads
GPT-5.4 mini: 42.9 (#107), Tulu 3 (Tülu 3) 70B: 37.1 (#217)
| Benchmark | GPT-5.4 mini | Tulu 3 (Tülu 3) 70B |
|---|---|---|
| LMArena Longer Query | 1407 | 1224 |
Writing & Preference GPT-5.4 mini leads
GPT-5.4 mini: 64.2 (#56), Tulu 3 (Tülu 3) 70B: 45.8 (#218)
| Benchmark | GPT-5.4 mini | Tulu 3 (Tülu 3) 70B |
|---|---|---|
| LMArena Text | 1411 | 1256 |
| LMArena Creative Writing | 1370 | 1231 |
| LMArena Multi-Turn | 1429 | 1252 |
| EQ-Bench Creative Writing | 1665 | — |
Frequently asked questions
Is GPT-5.4 mini better than Tulu 3 (Tülu 3) 70B?
GPT-5.4 mini is the stronger model overall, scoring 44.8 to 32.6 on the Noometry Index.
Is GPT-5.4 mini or Tulu 3 (Tülu 3) 70B better for coding?
GPT-5.4 mini scores higher on coding benchmarks: 45.0 versus 35.4 in the Noometry coding category.
How many benchmarks do GPT-5.4 mini and Tulu 3 (Tülu 3) 70B share?
13 benchmarks have published results for both models. GPT-5.4 mini has 46 scored results on Noometry and Tulu 3 (Tülu 3) 70B has 14.