Llama 3 70B Instruct
Apr 2024
Llama 3 70B InstructCompare inon
Benchmarks of Llama 3 70B Instruct16
BenchmarkByScoreRank
Legacy · saturated or no longer reported1
No benchmark matches
—per task?
API price not publicprice
—output speed?
8Kcontext
Ranking?
Alternatives?
ModelScoreCostPer taskWhat changes
BetterGPT-5 Minihighwins 60%—$0.053
OpenGemma 4 31Bunstatedwins 72%—$0.10 / $0.34BetterGPT-6 Lunamaxwins 71%—$0.068Function calling only with none · Chat Completions+3BetterGPT-5.6 Lunaxhighwins 71%—$0.085BetterGPT-5.6 Terraxhighwins 73%—$0.63Prompts with >272K input tokensModelScoreCostPer taskWhat changes
Nothing here yet.
ModelScoreCostPer taskWhat changes
GPT-5 Minihighwins 60%—$0.053GPT-6 Lunamaxwins 71%—$0.068Function calling only with none · Chat Completions+3GPT-5.6 Lunaxhighwins 71%—$0.085GPT-5.6 Terraxhighwins 73%—$0.63Prompts with >272K input tokensGPT-6.1 Solmaxwins 76%—$0.72Remove unsupported parameters+5ModelScoreCostPer taskWhat changes
Gemma 4 31BunstatedOpenwins 72%—$0.10 / $0.34
Mistral Medium 3.5unstatedOpenwins 70%—$1.50 / $7.50Memory (4-bit): 42 GB → 77 GBGemma 4 26B A4BunstatedOpenwins 70%—$0.15 / $0.60
Qwen3.8 27BunstatedOpenwins 66%—$0.42 / $2.55max_tokens range · QwenCloud+4Qwen 3.5 122B A10BunstatedOpenwins 65%—$0.40 / $3.20Memory (4-bit): 42 GB → 75 GBModelScoreCostPer taskWhat changes
Muse Wondermaxwins 79%—$4.52Only auto tool_choice is supported+5Muse Sparkunstatedwins 77%—$1.25 / $4.25Only n=1 · Meta / Chat Completions+1Muse Spark 1.3maxwins 77%—$1.60Only auto tool_choice is supported+5Muse Spark 1.2xhighwins 75%—$0.97Only auto tool_choice is supported+1Muse Spark 1.1unstatedwins 75%—$1.25 / $4.25Only n=1 · Meta / Chat Completions+1Muse Glimmer 30BunstatedOpenwins 65%—$0.30 / $1.10Llama 4 MaverickunstatedOpenwins 58%—$0.19 / $0.65Memory (4-bit): 42 GB → 241 GBLlama 4 ScoutunstatedOpenwins 55%—$0.10 / $0.30Memory (4-bit): 42 GB → 65 GBAll Meta models →Run it yourself
estimateFits on✗ 8 GB (phone, laptop)✗ 16 GB✗ 24 GB GPU✓ 48 GB GPU · compressed✓ 80 GB GPU · 8-bit
LicenceLlama 3 licencecommercial use with conditions
History
2 versionsBenchmarks
all categoriesChat & Knowledge
Arena Text · overall1,277#165
Arena Text · Chinese1,114.2#169
Arena Text · English1,277.3#157
Arena Text · French1,232.2#130
Arena Text · German1,169.1#137
Arena Text · instruction following1,194#164
Arena Text · Japanese1,016.5#131
Arena Text · Korean1,016.7#137
Arena Text · non-English1,141.7#171
Arena Text · Polish1,196.4#129
Arena Text · Russian1,158.7#168
Arena Text · Spanish1,241#137
Arena Text · overall · full cohort1,220.8#163
MMLU-Pro · official board56.2#12
Writing & Language
Arena Text · creative writing1,254#158
Arena Text · creative writing · full cohort1,209.6#157
FAQ
How much does Llama 3 70B Instruct cost?›
There is no per-unit API price: API price not public.
Is Llama 3 70B Instruct free to use?›
There is no free API. The weights are open, so you can run it yourself.
How good is Llama 3 70B Instruct, and what is it best at?›
#165 of 196 in Chat & Knowledge, #158 of 195 in Writing & Language.
Is Llama 3 70B Instruct better than Qwen2.5 72B Instruct?›
On Arena Text · creative writing, Llama 3 70B Instruct scores 1,254 vs 1,254 — about the same.
What is the context window of Llama 3 70B Instruct?›
8K tokens.
When was Llama 3 70B Instruct released? Is there a newer version?›
Released 2024-04-17. It is the current version in its line.
Is Llama 3 70B Instruct open source?›
Yes, the weights are open (llama3 licence).
Can I run Llama 3 70B Instruct locally?›
The weights are open and can be self-hosted; hardware needs are not confirmed yet.
Can Llama 3 70B Instruct read images?›
No, it accepts text only.
What is Llama 3 70B Instruct?›
Llama 3 70B Instruct is an open-weights language model from Meta, released Apr 2024.
What are the best alternatives to Llama 3 70B Instruct?›
GPT-5 Mini (better), Gemma 4 31B (open weights).