Model comparison
Llama-3.3-70B-Instruct vs Sonar
Sonar is the stronger model overall, scoring 37.8 to 30.3 on the Noometry Index. Llama-3.3-70B-Instruct costs 6.5× less per token, which makes it the better buy when Sonar's lead doesn't matter for your workload.
Last verified . 7 shared benchmarks.
Meta
PerplexitySummary
- They share 7 benchmarks with published results for both. Llama-3.3-70B-Instruct scores higher in 0 categories and Sonar in 5 categories; 4 gaps are clear of the uncertainty.
- The widest gap is in math, where Sonar leads 31.3 to 14.3.
- The biggest single-benchmark swing is LiveBench Data Analysis: 49.5% for Llama-3.3-70B-Instruct and 37.9% for Sonar.
- Llama-3.3-70B-Instruct is cheaper at $0.10 / $0.32 per million input/output tokens, against $1 / $1 for Sonar.
- Llama-3.3-70B-Instruct has downloadable open weights; the other is API-only.
Side by side
| Llama-3.3-70B-Instruct | Sonar | |
|---|---|---|
| Provider | Meta | Perplexity |
| Noometry Index | 30.3 | 37.8 |
| Released | 2024-12-06 | 2024-01-01 |
| Weights | Open | Proprietary |
| Context window | 128K | 128K |
| Max output | 4K | 4K |
| Input $ / M tokens | $0.10 | $1 |
| Output $ / M tokens | $0.32 | $1 |
| Results tracked | 43 | 7 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Sonar leads
Llama-3.3-70B-Instruct: 30.6 (#284), Sonar: 35.1 (#217)
| Benchmark | Llama-3.3-70B-Instruct | Sonar |
|---|---|---|
| LiveBench Coding | 36.6% | 35.1% |
| SciCode | 26% | — |
| WeirdML | 14.4% | — |
| BigCodeBench Instruct | 46.9% | — |
| LMArena Coding | 1268 | — |
| BigCodeBench Complete | 57.5% | — |
Agentic & Tool Use Not comparable
Llama-3.3-70B-Instruct: 25.6 (#100), Sonar: —
| Benchmark | Llama-3.3-70B-Instruct | Sonar |
|---|---|---|
| Berkeley Function Calling Leaderboard | 31.9% | — |
| BALROG | 23% | — |
Reasoning Sonar leads
Llama-3.3-70B-Instruct: 13.7 (#325), Sonar: 20.3 (#221)
| Benchmark | Llama-3.3-70B-Instruct | Sonar |
|---|---|---|
| LiveBench Reasoning | 50.8% | 46.3% |
| LiveBench Data Analysis | 49.5% | 37.9% |
| LiveBench | 50.2% | 46.9% |
| SimpleBench | 19.9% | — |
| CritPt | 0% | — |
| LMArena Hard Prompts | 1257 | — |
| DTBench | 59.5% | — |
| LMCA | 17.5% | — |
| Epoch Capabilities Index | 127.32 | — |
| ForecastBench | 58.6 | — |
Math Sonar leads
Llama-3.3-70B-Instruct: 14.3 (#292), Sonar: 31.3 (#197)
| Benchmark | Llama-3.3-70B-Instruct | Sonar |
|---|---|---|
| LiveBench Math | 42.2% | 41.6% |
| OTIS Mock AIME 2024-2025 | 5.1% | — |
| LMArena Math | 1268 | — |
| MATH Level 5 | 41.6% | — |
Knowledge Not comparable
Llama-3.3-70B-Instruct: 30.5 (#220), Sonar: —
| Benchmark | Llama-3.3-70B-Instruct | Sonar |
|---|---|---|
| GPQA Diamond | 47.4% | — |
| Confabulations | 22.8% | — |
| Vectara Hallucination Rate | 4.1% | — |
| LMArena Expert | 1225 | — |
| MMLU | 86.3% | — |
Multilingual Not comparable
Llama-3.3-70B-Instruct: 40.1 (#215), Sonar: —
| Benchmark | Llama-3.3-70B-Instruct | Sonar |
|---|---|---|
| LMArena Non-English | 1236 | — |
| LMArena Chinese | 1217 | — |
| LMArena French | 1280 | — |
| LMArena German | 1251 | — |
| LMArena Japanese | 1149 | — |
| LMArena Korean | 1142 | — |
| LMArena Russian | 1252 | — |
| LMArena Spanish | 1269 | — |
Instruction Following Too close to call
Llama-3.3-70B-Instruct: 71.3 (#152), Sonar: 71.5 (#146)
| Benchmark | Llama-3.3-70B-Instruct | Sonar |
|---|---|---|
| LiveBench Instruction Following | 82.7% | 76.2% |
| LMArena Instruction Following | 1242 | — |
Long Context Not comparable
Llama-3.3-70B-Instruct: 26.4 (#290), Sonar: —
| Benchmark | Llama-3.3-70B-Instruct | Sonar |
|---|---|---|
| Fiction.LiveBench | 33.3% | — |
| LMArena Longer Query | 1256 | — |
Writing & Preference Sonar leads
Llama-3.3-70B-Instruct: 47.8 (#203), Sonar: 52.6 (#166)
| Benchmark | Llama-3.3-70B-Instruct | Sonar |
|---|---|---|
| LiveBench Language | 39.2% | 44.1% |
| LMArena Text | 1274 | — |
| LMArena Creative Writing | 1250 | — |
| LMArena Multi-Turn | 1279 | — |
Frequently asked questions
Is Llama-3.3-70B-Instruct better than Sonar?
Sonar is the stronger model overall, scoring 37.8 to 30.3 on the Noometry Index. Llama-3.3-70B-Instruct costs 6.5× less per token, which makes it the better buy when Sonar's lead doesn't matter for your workload.
Which is cheaper, Llama-3.3-70B-Instruct or Sonar?
Llama-3.3-70B-Instruct is cheaper. It lists at $0.10 per million input tokens and $0.32 per million output tokens; Sonar lists at $1 and $1.
Is Llama-3.3-70B-Instruct or Sonar better for coding?
Sonar scores higher on coding benchmarks: 35.1 versus 30.6 in the Noometry coding category.
Which has the bigger context window?
Both accept 128K tokens.
How many benchmarks do Llama-3.3-70B-Instruct and Sonar share?
7 benchmarks have published results for both models. Llama-3.3-70B-Instruct has 43 scored results on Noometry and Sonar has 7.