MetaLlama 3.1 8B
Llama 3.1 8B is a Meta open-weights model, released 23 Jul 2024. Its best configuration ranks #753 of 785 on the BenchLeader Index at 36.1 ±4.1, in the lower half. The ± is the point: 151 other configurations score within that range, so they and this one cannot be told apart on quality alone — price and speed are what separate them. It scores highest in knowledge (48) and lowest in coding (19). At $0.048 per million tokens blended it is among the cheapest fifth of ranked models. Output speed of 136 tokens per second puts it faster than most, with a first answer in 0.9 s. Last measured 22 Oct 2026.
- Blended price
- $0.048/M
- $0.040 in · $0.070 out
- Output speed
- 136 tok/s
- measured by Artificial Analysis
- First answer
- 0.94 s
- first token 0.46 s
- Context
- 128k
- Full answer
- 4.61 s
- median, reasoning included
- Cost per run
- –
- Released
- 23 Jul 2024
- Overall index36
- Reasoning32
- Coding19
- Agents & tools37
- Maths31
- Knowledge48
- Instruction following34
- Human preference34
- Long context33
- Composite37
Versions
Meta has shipped 2 models under this name. Each is ranked on its own results; a newer version often has fewer results so far, which holds its index nearer the average until more arrive.
| Model | Released | Index | Rank |
|---|---|---|---|
| Llama 3.1 8Bthis page | 23 Jul 2024 | 36.1 | #753 |
| Llama 3-8B | 18 Apr 2024 | 35.1 | #764 |
How its index has moved
8 Sept 2026 to 23 Oct 2026The index is recomputed from scratch every day, so a line moves when a new benchmark result lands, when a publisher revises a score, or when the models it is normalised against change. Early movement usually means the score is still settling.
Benchmark results
See all LLM benchmarks. One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.
Reasoning
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| GPQA Diamond | 27.0%#280 | Epoch AI Benchmarking Hub | |
| Arena Hard Prompts | 1222#307 | Arena (LMArena) | |
| GPQA Diamond (AA)not in index | 25.9%#542 | Artificial Analysis | |
| Humanity's Last Exam (AA)not in index | 5.3%#436 | Artificial Analysis | |
| GPQA Diamond (HELM)not in index | 24.7%#66 | HELM Capabilities | |
| CritPt | 0.0%#315 | Artificial Analysis | |
| SciPredictnot in index | 14.7%#15 | Scale AI SEAL | |
| LMCAnot in index | 5.4%#189 | Epoch AI Benchmarking Hub | |
| DTBenchnot in index | 50.9%#211 | Epoch AI Benchmarking Hub | |
| PIQAnot in index | 81.2%#25 | Epoch AI Benchmarking Hub |
Coding
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| SciCode | 13.2%#214 | SciCode | |
| WeirdML | 1.7%#156 | WeirdML | |
| Arena Coding | 1259#302 | Arena (LMArena) | |
| Terminal-Bench 2.1 (AA)not in index | 1.5%#222 | Artificial Analysis |
Agents & tools
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Terminal-Bench Hard | 0.8%#349 | Artificial Analysis | |
| τ²-Bench Telecom (AA)not in index | 16.4%#354 | Artificial Analysis | |
| BFCL Overall | 25.8%#64 | Berkeley Function Calling Leaderboard | |
| BALROGnot in index | 15.1%#30 | BALROG |
Maths
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| OTIS Mock AIME | 1.7%#264 | Epoch AI Benchmarking Hub | |
| MATH Level 5 | 22.9%#76 | Epoch AI Benchmarking Hub | |
| Omni-MATH (HELM) | 13.7%#64 | HELM Capabilities | |
| Arena Maths | 1188#307 | Arena (LMArena) | |
| GSM8Knot in index | 82.4%#15 | Epoch AI Benchmarking Hub |
Knowledge
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| AA-Omniscience | -30.9#285 | Artificial Analysis | |
| MMLU-Pro (HELM)not in index | 40.6%#60 | HELM Capabilities | |
| MMLUnot in index | 56.1%#86 | Epoch AI Benchmarking Hub | |
| AA-Omniscience: accuracynot in index | 8.5%#514 | Artificial Analysis | |
| AA-Omniscience: non-hallucinationnot in index | 57.0%#87 | Artificial Analysis |
Instruction following
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| IFBench | 28.6%#372 | Artificial Analysis | |
| IFEval (HELM) | 74.3%#58 | HELM Capabilities | |
| WildBench (HELM) | 68.6%#60 | HELM Capabilities | |
| Arena Instruction Followingnot in index | 1191#307 | Arena (LMArena) |
Human preference
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Arena Text | 1211#306 | Arena (LMArena) | |
| Arena Creative Writingnot in index | 1177#306 | Arena (LMArena) | |
| Arena Multi-turnnot in index | 1198#299 | Arena (LMArena) | |
| Arena Longer Queriesnot in index | 1224#299 | Arena (LMArena) |
Long context
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| AA-LCR | 18.0%#444 | Artificial Analysis |
Composite
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Epoch Capabilities Indexnot in index | 116.5#189 | Epoch AI Benchmarking Hub | |
| AA Intelligence Index v4.3.2 | 6.9#503 | Artificial Analysis | |
| HELM Capabilities meannot in index | 44.4%#62 | HELM Capabilities |
Safety & honesty
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| PropensityBenchnot in index | 66.5%#5 | Scale AI SEAL |
Where to run it
Every provider serving this model through OpenRouter, with throughput and first-token latency measured on live traffic over the last 30 minutes and each provider’s own price. Purple marks the best in each column.
| Provider | Speed | First token | Input $/M | Output $/M | Context | Quantisation |
|---|---|---|---|---|---|---|
| Groq | 381 tok/s | 0.12 s | $0.050 | $0.080 | 131k | – |
| CoreWeave | 121 tok/s | 0.39 s | $0.220 | $0.220 | 131k | bf16 |
| NovitaAI | 96 tok/s | 0.46 s | $0.020 | $0.050 | 16k | fp8 |
| DeepInfra | 25 tok/s | 0.77 s | $0.020 | $0.040 | 131k | fp8 |
| Cloudflare | 16 tok/s | 0.38 s | $0.152 | $0.287 | 32k | fp8 |
Price history
The price above, per 1M tokens, every day it changed.
What a task costs
Estimates from list price, output speed and time to first answer for the best configuration. “With caching” assumes three-quarters of the input is served from the prompt cache. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0000 | – | 3.1 s |
| Summarize a 30-page report | 12,000 / 600 | $0.0005 | – | 5.3 s |
| Code edit | 6,000 / 1,500 | $0.0003 | – | 11.9 s |
| Agentic coding session | 60,000 / 4,000 | $0.0027 | – | 30.3 s |
| Structured extraction | 2,000 / 200 | $0.0001 | – | 2.4 s |
More from Meta
See also
Questions people ask
- What are Llama 3.1 8B's benchmark scores?
- As of 24 Oct 2026, Llama 3.1 8B scores 36.1 ±4.1 on the BenchLeader Index, strongest in knowledge (48) and weakest in coding (19). That rests on 19 independent benchmark results, and every score on this page links to its publisher.
- How does Llama 3.1 8B rank?
- Llama 3.1 8B is #753 of 785 ranked configurations on 24 Oct 2026. 151 other configurations score within its margin, so the exact position matters less than the band. The leader is Claude Opus 5.5 (max effort) at 72.6.
- How much does Llama 3.1 8B cost per 1M tokens?
- As of 24 Oct 2026, Llama 3.1 8B costs $0.040 per 1M input tokens and $0.070 per 1M output tokens, $0.048 blended (three input tokens per output token). The list price last changed on 13 Oct 2026.
- When was Llama 3.1 8B released?
- Meta released Llama 3.1 8B on 23 Jul 2024.
Data as of 24 Oct 2026. Compare with another model.
Cite as: BenchLeader, “Llama 3.1 8B: benchmarks, pricing, speed and rank”, data as of 24 Oct 2026, https://www.benchleader.com/models/llama-3-1-8b.