Llama 3.1 8B Instruct
Meta · released Jul 18, 2024 · meta-llama/Meta-Llama-3.1-8B-Instruct
- Type
- Open weightsLlama 3.1 Community License
- Params
- 8B
- Context
- 131K
about 98K words of context · download allowed, licence restricts use
Our take
Written Sep 17, 2026Llama 3.1 8B Instruct is a small text model you can download and run on a single modern graphics card, or reach cheaply through a host. Its licence puts conditions on commercial use and redistribution, and the measured quality here is weak, so treat it as a budget workhorse rather than a reasoning engine.
Use it for everyday chat and retrieval work where the bill matters more than peak quality, or when you want to run a model yourself on one modern graphics card. Its request capacity takes a long report or a stack of documents in one go, though reliable recall across all of it is unverified. Skip it if your task needs measured science or multi-task reasoning evidence, or if licence conditions on commercial use are a problem for you.
The case for it
- Small enough to run yourself: 8 billion parameters in total puts memory in use within reach of a single modern graphics card.
- Long documents need not be split up first, with a request capacity of 131072 tokens; room to hold them is not a guarantee of accurate recall.
- Cheap to try through a host, with several listed offers and the cheapest well under the rest.
The case against it
- Measured science and multi-task reasoning are weak: 2.5% correct on GPQA Diamond and 25.1% correct on MMLU-Pro, so tasks needing that knowledge need a trial on work you can check yourself.
- Instruction following is middling at 49.4% on IFEval, so precise multi-step instructions may need checking.
- The Llama 3.1 Community Licence puts conditions on commercial use and redistribution, so it needs reading before you build on it.
How good is it?
An open text model for chat and general instruction following, though it trails the field on everyday questions, writing and code.
- getting answers to everyday questionsArena Text (overall) · 166th of 168
- drafts, rewrites and editingArena Creative Writing · 166th of 168
- writing and completing codeArena Coding · 166th of 168
EverydayGeneral questions and everyday reasoning
Arena Text (overall)166th of 168 · 1211
CodingWriting and fixing code on its own
Arena Coding166th of 168 · 1260
Arena Coding is the only board that has scored it for this.
AgenticPlanning, calling tools, staying on task
Not yet scored on Arena Agent.
WritingDrafting and rewriting prose
Arena Creative Writing166th of 168 · 1177
Arena Creative Writing is the only board that has scored it for this.
These tests check whether a model follows instructions — a precondition for all the work above, but not a measure of how well that work is done.
Every published score for this model6 scoresEvery figure we hold, from 6 boards, with who ran it and a link to the source — including the boards no rating above is built on.
Can you run it yourself?
Comfortable fit
GeForce RTX 4090 · 24 GB
Room to spare. 16.3 GB spare means a 10% error in the size would not change the answer.
GeForce RTX 5090 · 32 GB
Room to spare. 24.3 GB spare means a 10% error in the size would not change the answer.
Apple M1 (8-core GPU) · 16 GB
Room to spare. 5.5 GB spare means a 10% error in the size would not change the answer.
Memory use by level
Against a 24 GB card.
What is quantisation? →This model on every device we track71 devicesThe Q4 build most people download, on each device: what the weights come to, how much context the memory leaves, and whether it runs. Smallest device that runs it first.
Check against your own machine → · Where to rent it hosted →
Or rent it from someone else
Prices checked 2 hours ago — each listing carries its own date.
Groq, through OpenRouter
Cheapest of the 2 listings we can compare like for like — at 131K of context, out of 6 in the table below. 2 cheaper rows there are outside that comparison: a different quantisation or a different context length.
- per 1M tokens
- $0.050 in / $0.080 out
- Context served
- 131K
- Throughput
- ~73 tok/s
| Provider | In / out per 1M tokens | Context | Throughput | Trains on prompts | Logs prompts | Zero retention |
|---|---|---|---|---|---|---|
| DeepInfrafp8Through OpenRouter | $0.020 / $0.040checked 2 hours ago | 131K16K max reply | 26 tok/s | No | No | Confirmed |
| Novita AIfp8Direct and through OpenRouter | $0.020 / $0.050checked 2 hours ago | 16K15K max reply through OpenRouter | 49 tok/sthrough OpenRouter | DirectUnknownThrough OpenRouterNo | DirectUnknownThrough OpenRouterNo | DirectUnknownThrough OpenRouterConfirmed |
| OpenRouterOpenRouter's own listing | $0.050 / $0.080checked 2 hours ago | 131K | not measured | Unknown | Unknown | Unknown |
| GroqThrough OpenRouter | $0.050 / $0.080checked 2 hours ago | 131K118K max reply | 73 tok/s | No | No | Confirmed |
| CoreWeavebf16Through OpenRouter | $0.22 / $0.22checked 2 hours ago | 131K118K max reply | 106 tok/s | No | No | Confirmed |
| Cloudflare Workers AIfp8Through OpenRouter | $0.15 / $0.29checked 2 hours ago | 32K29K max reply | 13 tok/s | No | Yesunknown period | Unknown |
DeepInfra
Novita AI
OpenRouter
Cloudflare Workers AIAcross the 6 listings we hold: 5 say they do not train on prompts (1 of them only through OpenRouter), 0 say they do and 1 does not say. 4 appear in the zero-retention registry we check (1 of them only through OpenRouter); the rest are unknown to us.
What each host's API supports
From the parameter list each endpoint publishes. Streaming is omitted: nothing we hold reports it, for any model.
| Provider | Tool calling | JSON output | Strict schema |
|---|---|---|---|
| DeepInfrafp8Through OpenRouter | ✗ | ✓ | ✗ |
| Novita AIfp8Direct and through OpenRouter | ✗ | ✓ | ✗ |
| OpenRouterOpenRouter's own listing | ✓ | ✓ | ✓ |
| GroqThrough OpenRouter | ✓ | ✓ | ✗ |
| CoreWeavebf16Through OpenRouter | ✓ | ✓ | ✓ |
| Cloudflare Workers AIfp8Through OpenRouter | ✗ | ✓ | ✗ |
Tool calling: 3 of 6 listings say yes, 3 say no. JSON output: 6 of 6 listings say yes. Strict schema: 2 of 6 listings say yes, 4 say no.
Models people weigh against Llama 3.1 8B Instruct
When we formed this view
Recent changes
What moved
first indexed by our pipelineEach date is the day we first saw the change, or the day the maker announced it.
What we do not know about this model yet
- We hold no measured file for it, so all 3 sizes on this page are calculated from the parameter count.
- Nothing we hold says whether an endpoint streams, so we do not show it either way.
- 1 of 6 listings does not say whether it trains on prompts, and 1 answers only through OpenRouter, not for its own listing.
- We hold no batch or off-peak rate for any of its listings.
- We hold a decode speed for it, but no prompt-processing (prefill) figure, so how long the input side of a job takes is unknown to us.
Licence and identifiers
What the licence allowsLlama 3.1 Community License, what it allows commercially, and the identifiers you need to pull this model — its Hugging Face repo, our slug and a machine-readable card.
Licence
Llama 3.1 Community License
Commercial use below 700M MAU. Notably allows using outputs to improve other models, which earlier Llama licenses banned. Derivative names must start with "Llama".
Identifiers
- Hugging Face
- meta-llama/Meta-Llama-3.1-8B-Instruct
- Takes in, gives back
- Text in, text out
- Catalogue slug
- meta-llama-llama-3-1-8b-instruct