Model comparison
GPT-5.4 mini vs Inkling-Small
Inkling-Small is the stronger model overall, scoring 46.5 to 44.8 on the Noometry Index.
Last verified . 33 shared benchmarks.
Summary
- They share 33 benchmarks with published results for both. GPT-5.4 mini scores higher in 8 categories and Inkling-Small in 1 category; 4 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where Inkling-Small leads 39.0 to 30.3.
- The biggest single-benchmark swing is ARC-AGI-2: 18.9% for GPT-5.4 mini and 40.1% for Inkling-Small.
- Inkling-Small is cheaper at $0.45 / $1.20 per million input/output tokens, against $0.75 / $4.50 for GPT-5.4 mini.
- Inkling-Small accepts more context: 524K tokens versus 400K.
- Inkling-Small has downloadable open weights; the other is API-only.
Side by side
| GPT-5.4 mini | Inkling-Small | |
|---|---|---|
| Provider | OpenAI | Thinking Machines Lab |
| Noometry Index | 44.8 | 46.5 |
| Released | 2026-03-17 | 2026-07-15 |
| Weights | Proprietary | Open |
| Context window | 400K | 524K |
| Max output | 128K | 1.05M |
| Input $ / M tokens | $0.75 | $0.45 |
| Output $ / M tokens | $4.50 | $1.20 |
| Results tracked | 46 | 33 |
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), Inkling-Small: 43.3 (#80)
| Benchmark | GPT-5.4 mini | Inkling-Small |
|---|---|---|
| LMArena WebDev | 1397 | 1408 |
| SciCode | 49.9% | 48.7% |
| LMArena Coding | 1438 | 1451 |
| FrontierCode | 27% | — |
| WeirdML | 60.3% | — |
| ALE-Bench | 1,189 | — |
Agentic & Tool Use Not comparable
GPT-5.4 mini: 29.6 (#74), Inkling-Small: —
| Benchmark | GPT-5.4 mini | Inkling-Small |
|---|---|---|
| DeepResearch Bench | 36.3% | — |
Reasoning Inkling-Small leads
GPT-5.4 mini: 30.3 (#77), Inkling-Small: 39.0 (#53)
| Benchmark | GPT-5.4 mini | Inkling-Small |
|---|---|---|
| ARC-AGI-2 | 18.9% | 40.1% |
| ARC-AGI-1 | 63.7% | 84% |
| CritPt | 10% | 8.3% |
| Chess Puzzles | 24% | 18% |
| LMArena Hard Prompts | 1424 | 1424 |
| Mystery Game Puzzles | 11% | 6% |
| Epoch Capabilities Index | 148.83 | 150.13 |
| Kagi LLM Benchmark | 37.9% | — |
| NYT Connections (extended) | 61.8% | — |
| Thematic Generalization | 61.7% | — |
| DTBench | 80% | — |
| LMCA | 40.8% | — |
| ForecastBench | 57 | — |
Math Too close to call
GPT-5.4 mini: 45.4 (#63), Inkling-Small: 44.9 (#65)
| Benchmark | GPT-5.4 mini | Inkling-Small |
|---|---|---|
| FrontierMath (Tiers 1-3) | 51.2% | 46.3% |
| FrontierMath Tier 4 | 9.8% | 17.1% |
| OTIS Mock AIME 2024-2025 | 88.9% | 90% |
| ProofBench | 21% | 6% |
| LMArena Math | 1418 | 1453 |
| 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), Inkling-Small: 48.3 (#71)
| Benchmark | GPT-5.4 mini | Inkling-Small |
|---|---|---|
| GPQA Diamond | 86.9% | 88.5% |
| SimpleQA Verified | 29.4% | 19.1% |
| LMArena Expert | 1434 | 1442 |
| Vectara Hallucination Rate | 5.5% | — |
Multimodal Too close to call
GPT-5.4 mini: 39.0 (#51), Inkling-Small: 38.4 (#59)
| Benchmark | GPT-5.4 mini | Inkling-Small |
|---|---|---|
| LMArena Vision | 1245 | 1235 |
Multilingual Too close to call
GPT-5.4 mini: 52.1 (#94), Inkling-Small: 51.7 (#102)
| Benchmark | GPT-5.4 mini | Inkling-Small |
|---|---|---|
| LMArena Non-English | 1405 | 1400 |
| LMArena Chinese | 1446 | 1458 |
| LMArena French | 1440 | 1438 |
| LMArena German | 1407 | 1412 |
| LMArena Japanese | 1375 | 1399 |
| LMArena Korean | 1369 | 1355 |
| LMArena Russian | 1417 | 1392 |
| LMArena Spanish | 1403 | 1426 |
Instruction Following Too close to call
GPT-5.4 mini: 74.3 (#99), Inkling-Small: 73.9 (#112)
| Benchmark | GPT-5.4 mini | Inkling-Small |
|---|---|---|
| LMArena Instruction Following | 1405 | 1397 |
Long Context Too close to call
GPT-5.4 mini: 42.9 (#107), Inkling-Small: 42.7 (#113)
| Benchmark | GPT-5.4 mini | Inkling-Small |
|---|---|---|
| LMArena Longer Query | 1407 | 1401 |
Writing & Preference GPT-5.4 mini leads
GPT-5.4 mini: 64.2 (#56), Inkling-Small: 59.8 (#102)
| Benchmark | GPT-5.4 mini | Inkling-Small |
|---|---|---|
| LMArena Text | 1411 | 1413 |
| LMArena Creative Writing | 1370 | 1330 |
| EQ-Bench Creative Writing | 1665 | 1491 |
| LMArena Multi-Turn | 1429 | 1420 |
Frequently asked questions
Is GPT-5.4 mini better than Inkling-Small?
Inkling-Small is the stronger model overall, scoring 46.5 to 44.8 on the Noometry Index.
Which is cheaper, GPT-5.4 mini or Inkling-Small?
Inkling-Small is cheaper. It lists at $0.45 per million input tokens and $1.20 per million output tokens; GPT-5.4 mini lists at $0.75 and $4.50.
Is GPT-5.4 mini or Inkling-Small better for coding?
GPT-5.4 mini scores higher on coding benchmarks: 45.0 versus 43.3 in the Noometry coding category.
Which has the bigger context window?
Inkling-Small does, with 524K tokens against 400K.
How many benchmarks do GPT-5.4 mini and Inkling-Small share?
33 benchmarks have published results for both models. GPT-5.4 mini has 46 scored results on Noometry and Inkling-Small has 33.