# Inkling-Small vs Llama 3.1-70B

> Inkling-Small is the stronger model overall, scoring 46.5 to 29.4 on the Noometry Index. Llama 3.1-70B costs 1.6× less per token, which makes it the better buy when Inkling-Small's lead doesn't matter for your workload.

- Canonical page: https://noometry.com/compare/inkling-small-vs-llama-3-1-70b
- Last updated: 2026-10-10
- Shared benchmarks: 21

## Summary

- They share 21 benchmarks with published results for both. Inkling-Small scores higher in 8 categories and Llama 3.1-70B in 0 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in math, where Inkling-Small leads 44.9 to 12.7.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 90% for Inkling-Small and 3.6% for Llama 3.1-70B.
- Llama 3.1-70B is cheaper at $0.40 / $0.40 per million input/output tokens, against $0.45 / $1.20 for Inkling-Small.
- Inkling-Small accepts more context: 524K tokens versus 128K.

## Snapshot

| | Inkling-Small | Llama 3.1-70B |
|---|---|---|
| Provider | Thinking Machines Lab | Meta |
| Noometry Index | 46.5 | 29.4 |
| Rank | 57 | 299 |
| Context | 524K | 128K |
| Input $/M | $0.45 | $0.40 |
| Output $/M | $1.20 | $0.40 |
| Weights | Open | Open |

## Coding

- Inkling-Small: 43.3 (#80)
- Llama 3.1-70B: 29.8 (#292)

| Benchmark | Inkling-Small | Llama 3.1-70B |
|---|---|---|
| LMArena Coding | 1451 | 1260 |
| LMArena WebDev | 1408 | — |
| SciCode | 48.7% | — |
| WeirdML | — | 9% |
| BigCodeBench Instruct | — | 46.1% |
| BigCodeBench Complete | — | 54.8% |

## Agentic & Tool Use

- Inkling-Small: —
- Llama 3.1-70B: 24.9 (#104)

| Benchmark | Inkling-Small | Llama 3.1-70B |
|---|---|---|
| TheAgentCompany | — | 6.9% |
| BALROG | — | 27.9% |

## Reasoning

- Inkling-Small: 39.0 (#53)
- Llama 3.1-70B: 20.8 (#215)

| Benchmark | Inkling-Small | Llama 3.1-70B |
|---|---|---|
| LMArena Hard Prompts | 1424 | 1241 |
| Epoch Capabilities Index | 150.13 | 125.91 |
| ARC-AGI-2 | 40.1% | — |
| ARC-AGI-1 | 84% | — |
| CritPt | 8.3% | — |
| Chess Puzzles | 18% | — |
| Mystery Game Puzzles | 6% | — |
| DTBench | — | 60% |
| LMCA | — | 14.8% |

## Math

- Inkling-Small: 44.9 (#65)
- Llama 3.1-70B: 12.7 (#298)

| Benchmark | Inkling-Small | Llama 3.1-70B |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 90% | 3.6% |
| LMArena Math | 1453 | 1252 |
| FrontierMath (Tiers 1-3) | 46.3% | — |
| FrontierMath Tier 4 | 17.1% | — |
| ProofBench | 6% | — |
| Omni-MATH | — | 21% |
| MATH Level 5 | — | 36.7% |

## Knowledge

- Inkling-Small: 48.3 (#71)
- Llama 3.1-70B: 24.1 (#264)

| Benchmark | Inkling-Small | Llama 3.1-70B |
|---|---|---|
| GPQA Diamond | 88.5% | 44.2% |
| LMArena Expert | 1442 | 1209 |
| SimpleQA Verified | 19.1% | — |
| MMLU-Pro | — | 65.3% |
| GPQA (HELM) | — | 42.6% |
| MMLU | — | 80.1% |

## Multimodal

- Inkling-Small: 38.4 (#59)
- Llama 3.1-70B: —

| Benchmark | Inkling-Small | Llama 3.1-70B |
|---|---|---|
| LMArena Vision | 1235 | — |

## Multilingual

- Inkling-Small: 51.7 (#102)
- Llama 3.1-70B: 38.9 (#220)

| Benchmark | Inkling-Small | Llama 3.1-70B |
|---|---|---|
| LMArena Non-English | 1400 | 1219 |
| LMArena Chinese | 1458 | 1215 |
| LMArena French | 1438 | 1260 |
| LMArena German | 1412 | 1222 |
| LMArena Japanese | 1399 | 1132 |
| LMArena Korean | 1355 | 1140 |
| LMArena Russian | 1392 | 1234 |
| LMArena Spanish | 1426 | 1253 |

## Instruction Following

- Inkling-Small: 73.9 (#112)
- Llama 3.1-70B: 65.5 (#218)

| Benchmark | Inkling-Small | Llama 3.1-70B |
|---|---|---|
| LMArena Instruction Following | 1397 | 1231 |
| IFEval | — | 82.1% |

## Long Context

- Inkling-Small: 42.7 (#113)
- Llama 3.1-70B: 37.6 (#208)

| Benchmark | Inkling-Small | Llama 3.1-70B |
|---|---|---|
| LMArena Longer Query | 1401 | 1241 |

## Writing & Preference

- Inkling-Small: 59.8 (#102)
- Llama 3.1-70B: 35.6 (#262)

| Benchmark | Inkling-Small | Llama 3.1-70B |
|---|---|---|
| LMArena Text | 1413 | 1261 |
| LMArena Creative Writing | 1330 | 1232 |
| EQ-Bench Creative Writing | 1491 | 784 |
| LMArena Multi-Turn | 1420 | 1256 |
| WildBench | — | 75.8% |

## FAQ

### Is Inkling-Small better than Llama 3.1-70B?

Inkling-Small is the stronger model overall, scoring 46.5 to 29.4 on the Noometry Index. Llama 3.1-70B costs 1.6× less per token, which makes it the better buy when Inkling-Small's lead doesn't matter for your workload.

### Which is cheaper, Inkling-Small or Llama 3.1-70B?

Llama 3.1-70B is cheaper. It lists at $0.40 per million input tokens and $0.40 per million output tokens; Inkling-Small lists at $0.45 and $1.20.

### Is Inkling-Small or Llama 3.1-70B better for coding?

Inkling-Small scores higher on coding benchmarks: 43.3 versus 29.8 in the Noometry coding category.

### Which has the bigger context window?

Inkling-Small does, with 524K tokens against 128K.

### How many benchmarks do Inkling-Small and Llama 3.1-70B share?

21 benchmarks have published results for both models. Inkling-Small has 33 scored results on Noometry and Llama 3.1-70B has 35.