GPT-5.4 Mini vs Llama 3.3 70B Instruct
GPT-5.4 Mini leads Llama 3.3 70B Instruct on the Artificial Analysis Intelligence Index, 24.1 to 7.7, and Llama 3.3 70B Instruct costs 6.3x less per output token ($0.72 vs $4.50 per million). Updated October 21, 2026; 42 benchmarks have a result for both.
GPT-5.4 Mini or Llama 3.3 70B Instruct: which to pick
Only GPT-5.4 Mini has a coding score, 35.8 on the Vibe Coding Index; Llama 3.3 70B Instruct has none, so the two cannot be ranked on coding.
Pick Llama 3.3 70B Instruct for output tokens 6.3x cheaper ($0.72 vs $4.50 per million).
- Shared results
- 42
- GPT-5.4 Mini only
- 47
- Llama 3.3 70B Instruct only
- 10
- Like-for-like categories
- 7 of 9
Only benchmarks that tested both models count toward a pick; a result only one model has is shown but not compared.
Which one for your work
Each card is decided by one number both models carry. Where only one has it, the card says so and makes no call.
Coding
Vibe Coding Index, six public coding benchmarks
No call
Only GPT-5.4 Mini has a coding score, so no call is made.
One side not measured
Intelligence
AA Intelligence Index
GPT-5.4 Mini
GPT-5.4 Mini leads, 24.1 against 7.7.
Same board, both measured
Agentic work
AA Agentic Index
No call
Only GPT-5.4 Mini has a AA Agentic Index result, 17.9.
Directional only
Reasoning
GPQA Diamond
No call
Only GPT-5.4 Mini has a GPQA Diamond result, 83.1%.
Directional only
Math
AIME
No call
Only GPT-5.4 Mini has a AIME result, 95.6%.
Directional only
Knowledge
MMLU Pro
No call
Only GPT-5.4 Mini has a MMLU Pro result, 84.5%.
Directional only
Long documents
Listed context window
GPT-5.4 Mini
GPT-5.4 Mini takes 400K tokens against 131K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
Llama 3.3 70B Instruct
GPT-5.4 Mini $0.0030, Llama 3.3 70B Instruct $0.0008 per turn.
List prices
Cost per solved task
One SWE-bench Verified task, solved
No call
Only GPT-5.4 Mini has the number, $0.70.
Directional only
At a glance
- Coding index
- GPT-5.4 Mini 35.8 from 5 benchmarks, Llama 3.3 70B Instruct not benchmarked.
- Widest category
- Agentic work, 56.7 points apart on average over 1 shared benchmark.
- Widest single gap
- OTIS Mock AIME: GPT-5.4 Mini 88.9% against Llama 3.3 70B Instruct 5.1%, 83.8 points apart on a 0 to 100 scale.
- Output price
- GPT-5.4 Mini $4.50 against Llama 3.3 70B Instruct $0.72 per million output tokens, 6.3x apart.
- Context window
- GPT-5.4 Mini takes 400K tokens, Llama 3.3 70B Instruct 131K.
Benchmark by benchmark
First the score: boards each model leads, per category, where both have a result. Then every board, a category to a fold; bars are each result on a 0 to 100 scale, so compare the two in a row, not one row with another.
| Category | GPT-5.4 Mini | Bars | Llama 3.3 70B Instruct | Ties |
|---|---|---|---|---|
| Coding | 5 | 0 | ||
| Intelligence | 2 | 0 | ||
| Agentic work | 1 | 0 | ||
| Reasoning | 6 | 0 | ||
| Math | 3 | 0 | ||
| Knowledge | 4 | 0 | ||
| General | 21 | 0 | ||
| All shared boards | 42 | 0 |
CodingLike-for-likeSix of these boards feed the Vibe Coding Index50
| Benchmark | GPT-5.4 Mini | Llama 3.3 70B Instruct | Leader |
|---|---|---|---|
| LMArena WebDev | 1397 | Not measured | Only GPT-5.4 Mini measured |
| AA Coding Index | 56.1 | 11.9 | GPT-5.4 Mini |
| LiveCodeBench (Vals AI) | 81.5% | Not measured | Only GPT-5.4 Mini measured |
| SWE-bench Verified (Vals AI) | 73.0% | Not measured | Only GPT-5.4 Mini measured |
| Terminal-Bench 4 (Vals AI) | 2.5% | Not measured | Only GPT-5.4 Mini measured |
| Vibe Code Bench (Vals AI) | 48.0% | Not measured | Only GPT-5.4 Mini measured |
| Code migration (Vals AI) | 12.9% | Not measured | Only GPT-5.4 Mini measured |
| ProgramBench (Vals AI) | 0.0% | Not measured | Only GPT-5.4 Mini measured |
| Terminal-Bench 2.1 (Vals AI) | 54.7% | Not measured | Only GPT-5.4 Mini measured |
| LiveCodeBench (Artificial Analysis run) | Not measured | 28.8% | Only Llama 3.3 70B Instruct measured |
| SciCode | 52.1% | Not measured | Only GPT-5.4 Mini measured |
| Terminal-Bench Hard (Artificial Analysis run) | 52.3% | 3.0% | GPT-5.4 Mini |
| FrontierCode | 27.0% | Not measured | Only GPT-5.4 Mini measured |
| WeirdML | 60.3% | 14.4% | GPT-5.4 Mini |
| ALE-Bench | 1188.6 | Not measured | Only GPT-5.4 Mini measured |
| LMArena Text: coding | 1438 | 1268 | GPT-5.4 Mini |
| LMArena Text: software and IT | 1436 | 1273 | GPT-5.4 Mini |
| LMArena WebDev: React | 1386 | Not measured | Only GPT-5.4 Mini measured |
| LMArena WebDev: HTML | 1419 | Not measured | Only GPT-5.4 Mini measured |
| Neither model has a result on the other 20 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index20
| Benchmark | GPT-5.4 Mini | Llama 3.3 70B Instruct | Leader |
|---|---|---|---|
| AA Intelligence Index | 24.1 | 7.7 | GPT-5.4 Mini |
| Epoch Capabilities Index | 148.7 | 127.3 | GPT-5.4 Mini |
Agentic workLike-for-likePublic boards, shown as published, kept out of the index10
| Benchmark | GPT-5.4 Mini | Llama 3.3 70B Instruct | Leader |
|---|---|---|---|
| AA Agentic Index | 17.9 | Not measured | Only GPT-5.4 Mini measured |
| Finance Agent (Vals AI) | 45.4% | Not measured | Only GPT-5.4 Mini measured |
| Tax Agent Bench (Vals AI) | 8.8% | Not measured | Only GPT-5.4 Mini measured |
| Legal Research Bench (Vals AI) | 12.5% | Not measured | Only GPT-5.4 Mini measured |
| Excel Modeling Benchmark (Vals AI) | 45.4% | Not measured | Only GPT-5.4 Mini measured |
| τ²-Bench Telecom | 83.3% | 26.6% | GPT-5.4 Mini |
| BALROG | Not measured | 23.0% | Only Llama 3.3 70B Instruct measured |
| DeepResearch Bench | 36.3% | Not measured | Only GPT-5.4 Mini measured |
| Berkeley Function Calling Leaderboard (BFCL) | Not measured | 31.9% | Only Llama 3.3 70B Instruct measured |
| Neither model has a result on the other 32 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index60
| Benchmark | GPT-5.4 Mini | Llama 3.3 70B Instruct | Leader |
|---|---|---|---|
| GPQA Diamond | 83.1% | Not measured | Only GPT-5.4 Mini measured |
| MMMU Pro (Vals AI) | 79.3% | Not measured | Only GPT-5.4 Mini measured |
| GPQA Diamond (Artificial Analysis run) | 87.5% | 49.8% | GPT-5.4 Mini |
| Humanity's Last Exam (Artificial Analysis run) | 28.1% | 3.6% | GPT-5.4 Mini |
| GPQA Diamond (Epoch AI run) | 86.9% | 47.4% | GPT-5.4 Mini |
| ARC-AGI | 63.7% | Not measured | Only GPT-5.4 Mini measured |
| ARC-AGI-2 | 18.9% | Not measured | Only GPT-5.4 Mini measured |
| SimpleBench | Not measured | 19.9% | Only Llama 3.3 70B Instruct measured |
| Chess puzzles | 23.0% | Not measured | Only GPT-5.4 Mini measured |
| Mystery game puzzles | 11.0% | Not measured | Only GPT-5.4 Mini measured |
| LMArena Text: hard prompts | 1424 | 1257 | GPT-5.4 Mini |
| LMArena Text: expert | 1435 | 1225 | GPT-5.4 Mini |
| LMArena Text: hard prompts in English | 1420 | 1277 | GPT-5.4 Mini |
| LMArena Vision: diagrams | 1266 | Not measured | Only GPT-5.4 Mini measured |
| LMArena Vision: homework | 1285 | Not measured | Only GPT-5.4 Mini measured |
| Kagi LLM Benchmark | 37.9% | Not measured | Only GPT-5.4 Mini measured |
| Neither model has a result on the other 6 boards here | |||
MathLike-for-likePublic boards, shown as published, kept out of the index30
| Benchmark | GPT-5.4 Mini | Llama 3.3 70B Instruct | Leader |
|---|---|---|---|
| AIME | 95.6% | Not measured | Only GPT-5.4 Mini measured |
| AA Math Index | Not measured | 7.7 | Only Llama 3.3 70B Instruct measured |
| AIME 2025 (Artificial Analysis run) | Not measured | 7.7% | Only Llama 3.3 70B Instruct measured |
| OTIS Mock AIME | 88.9% | 5.1% | GPT-5.4 Mini |
| FrontierMath, tiers 1 to 3 | 51.2% | Not measured | Only GPT-5.4 Mini measured |
| FrontierMath, tier 4 | 9.8% | Not measured | Only GPT-5.4 Mini measured |
| ProofBench | 21.0% | Not measured | Only GPT-5.4 Mini measured |
| MATH level 5 | Not measured | 41.6% | Only Llama 3.3 70B Instruct measured |
| LMArena Text: math | 1418 | 1267 | GPT-5.4 Mini |
| LMArena Text: mathematical work | 1429 | 1267 | GPT-5.4 Mini |
| Neither model has a result on the other 3 boards here | |||
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index40
| Benchmark | GPT-5.4 Mini | Llama 3.3 70B Instruct | Leader |
|---|---|---|---|
| MMLU Pro | 84.5% | Not measured | Only GPT-5.4 Mini measured |
| CaseLaw (Vals AI) | 51.7% | Not measured | Only GPT-5.4 Mini measured |
| CorpFin (Vals AI) | 60.9% | Not measured | Only GPT-5.4 Mini measured |
| TaxEval (Vals AI) | 71.2% | Not measured | Only GPT-5.4 Mini measured |
| MortgageTax (Vals AI) | 63.5% | Not measured | Only GPT-5.4 Mini measured |
| MMLU Pro (Artificial Analysis run) | Not measured | 71.3% | Only Llama 3.3 70B Instruct measured |
| SimpleQA Verified | 29.4% | Not measured | Only GPT-5.4 Mini measured |
| MMLU | Not measured | 86.3% | Only Llama 3.3 70B Instruct measured |
| LMArena Text: science | 1417 | 1284 | GPT-5.4 Mini |
| LMArena Text: medicine and healthcare | 1395 | 1275 | GPT-5.4 Mini |
| LMArena Text: legal and government | 1414 | 1284 | GPT-5.4 Mini |
| LMArena Text: business and finance | 1418 | 1243 | GPT-5.4 Mini |
| Neither model has a result on the other 6 boards here | |||
Web designNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | GPT-5.4 Mini | Llama 3.3 70B Instruct | Leader |
|---|---|---|---|
| Neither model has a result on 9 boards here | |||
GeneralLike-for-likePublic boards, shown as published, kept out of the index210
| Benchmark | GPT-5.4 Mini | Llama 3.3 70B Instruct | Leader |
|---|---|---|---|
| SAGE (Vals AI) | 50.8% | Not measured | Only GPT-5.4 Mini measured |
| Vals Multimodal Index | 54.2% | Not measured | Only GPT-5.4 Mini measured |
| IFBench | 73.3% | 47.1% | GPT-5.4 Mini |
| Long context reasoning (AA-LCR) | 77.0% | 15.7% | GPT-5.4 Mini |
| DTBench | 80.0% | 59.5% | GPT-5.4 Mini |
| LMCA | 40.8% | 17.5% | GPT-5.4 Mini |
| Fiction.LiveBench | Not measured | 33.3% | Only Llama 3.3 70B Instruct measured |
| LMArena Text | 1412 | 1274 | GPT-5.4 Mini |
| LMArena Text: instruction following | 1405 | 1242 | GPT-5.4 Mini |
| LMArena Text: creative writing | 1370 | 1250 | GPT-5.4 Mini |
| LMArena Vision | 1245 | Not measured | Only GPT-5.4 Mini measured |
| LMArena Text: English | 1412 | 1303 | GPT-5.4 Mini |
| LMArena Text: non-English | 1405 | 1236 | GPT-5.4 Mini |
| LMArena Text: Chinese | 1446 | 1217 | GPT-5.4 Mini |
| LMArena Text: German | 1408 | 1251 | GPT-5.4 Mini |
| LMArena Text: Japanese | 1375 | 1150 | GPT-5.4 Mini |
| LMArena Text: Korean | 1370 | 1142 | GPT-5.4 Mini |
| LMArena Text: French | 1441 | 1282 | GPT-5.4 Mini |
| LMArena Text: Spanish | 1405 | 1269 | GPT-5.4 Mini |
| LMArena Text: Russian | 1417 | 1252 | GPT-5.4 Mini |
| LMArena Text: Polish | 1415 | 1259 | GPT-5.4 Mini |
| LMArena Text: multi-turn | 1429 | 1280 | GPT-5.4 Mini |
| LMArena Text: longer queries | 1408 | 1255 | GPT-5.4 Mini |
| LMArena Text: writing and language | 1393 | 1255 | GPT-5.4 Mini |
| LMArena Text: entertainment and media | 1373 | 1241 | GPT-5.4 Mini |
| LMArena Vision: English | 1244 | Not measured | Only GPT-5.4 Mini measured |
| LMArena Vision: Chinese | 1286 | Not measured | Only GPT-5.4 Mini measured |
| LMArena Vision: reading text | 1253 | Not measured | Only GPT-5.4 Mini measured |
| LMArena Vision: humor | 1235 | Not measured | Only GPT-5.4 Mini measured |
| LMArena Vision: recognizing things | 1218 | Not measured | Only GPT-5.4 Mini measured |
| LMArena Vision: creative writing | 1222 | Not measured | Only GPT-5.4 Mini measured |
| Neither model has a result on the other 15 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | GPT-5.4 Mini | Llama 3.3 70B Instruct | Leader |
|---|---|---|---|
| Neither model has a result on 5 boards here | |||
What each workload costs
Three fixed token mixes at each model's listed price, so per-token rates become a bill.
Chat turn
1K input + 500 output tokens
Llama 3.3 70B Instruct costs less
Repository review
50K input + 3K output tokens
Llama 3.3 70B Instruct costs less
Agent loop
200K input + 10K output tokens
Llama 3.3 70B Instruct costs less
Listed OpenRouter rates, without cache or batch discounts. A workload that does not fit a model's context window would need splitting.
Specification differences
The leading value in a row is marked; a row with a missing value names no leader. Our two rows are marked in red.
| Specification | OpenAI GPT-5.4 Mini | Meta Llama 3.3 70B Instruct |
|---|---|---|
| Maker | OpenAI | Meta Llama |
| Weights | No public weights listed | Open weights |
| Released | March 17, 2026 | December 6, 2024 |
| Context window | 400K | 131K |
| Input $/Mtok | $0.75 | $0.45 |
| Output $/Mtok | $4.50 | $0.72 |
| Coding index | 35.8 | Not benchmarked |
| Independent measurements behind it | 5 | 0 |
| Cost per solved task | $0.70 | Not computable |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is GPT-5.4 Mini better than Llama 3.3 70B Instruct for coding?
Only GPT-5.4 Mini has a coding score, 35.8 on the Vibe Coding Index; Llama 3.3 70B Instruct has none, so the two cannot be ranked on coding. Pick Llama 3.3 70B Instruct for output tokens 6.3x cheaper ($0.72 vs $4.50 per million).
Which is cheaper, GPT-5.4 Mini or Llama 3.3 70B Instruct?
GPT-5.4 Mini costs $0.75 per million input tokens and $4.50 per million output tokens; Llama 3.3 70B Instruct costs $0.45 and $0.72. Cost per solved task is known for GPT-5.4 Mini only.
Which has the bigger context window, GPT-5.4 Mini or Llama 3.3 70B Instruct?
GPT-5.4 Mini, with 400K tokens against 131K.
How many benchmarks is the GPT-5.4 Mini vs Llama 3.3 70B Instruct comparison based on?
42 benchmarks have a result for both models as of October 21, 2026. GPT-5.4 Mini's coding index rests on 5 independent benchmarks and Llama 3.3 70B Instruct's on 0. Where only one model has a result, the page says so instead of naming a winner.