Llama 3.1 70B Instruct
Llama 3.1 70B InstructCompare inon
Benchmarks of Llama 3.1 70B Instruct19
BenchmarkByScoreRank
Math1
Writing15
Legacy · saturated or no longer reported3
No benchmark matches
Rankingconfiguration unstated?
Alternatives?
ModelScoreCostPer taskWhat changes
CheaperGemma 4 31BunstatedOpen+75%-60%$0.10 / $0.34
BetterGPT-6 Astraunstated×2.8+4,900%$10 / $50Monitoring can return HTTP 403+2OpenMuse Glimmer 30Bunstated×2.3+25%$0.30 / $1.10CheaperGemma 4 26B A4BunstatedOpen+66%-34%$0.15 / $0.60CheaperGemma 4 12BunstatedOpen+64%-63%$0.10 / $0.30ModelScoreCostPer taskWhat changes
Gemma 4 31BunstatedOpen+75%-60%$0.10 / $0.34Gemma 4 26B A4BunstatedOpen+66%-34%$0.15 / $0.60Gemma 4 12BunstatedOpen+64%-63%$0.10 / $0.30Gemma 3 27BunstatedOpen+62%-75%$0.08 / $0.16
QwQ 32BOpen+60%-47%$0.15 / $0.40ModelScoreCostPer taskWhat changes
GPT-6 Astraunstated×2.8+4,900%$10 / $50Monitoring can return HTTP 403+2
Claude Fable 5.1unstated×2.8+4,900%$10 / $50Replace forced tool_choice+5Claude Opus 5unstated×2.7+2,400%$5 / $25GPT-6 Solunstated×2.7+900%$2 / $10
Kimi K3unstatedOpen×2.7+1,400%$3 / $15Memory (4-bit): 42 GB → 1,668 GB+4ModelScoreCostPer taskWhat changes
Muse Glimmer 30BunstatedOpen×2.3+25%$0.30 / $1.10Qwen3.8 27BunstatedOpen×2.1+139%$0.42 / $2.55max_tokens range · QwenCloud+4Gemma 4 31BunstatedOpen+75%-60%$0.10 / $0.34Gemma 4 26B A4BunstatedOpen+66%-34%$0.15 / $0.60Gemma 4 12BunstatedOpen+64%-63%$0.10 / $0.30ModelScoreCostPer taskWhat changes
Muse Spark 1.1unstated×2.5—$1.25 / $4.25Only n=1 · Meta / Chat Completions+1Muse Spark 1.3unstated×2.4—$1.25 / $4.25Only auto tool_choice is supported+5Muse Spark 1.2unstated×2.3—$1.25 / $4.25Only auto tool_choice is supported+1Muse Glimmer 30BunstatedOpen×2.3—$0.30 / $1.10Llama 4 MaverickunstatedOpen+10%—$0.19 / $0.65Memory (4-bit): 42 GB → 241 GBLlama 4 ScoutunstatedOpensame—$0.10 / $0.30Memory (4-bit): 42 GB → 65 GBLlama 3.1 8B InstructunstatedOpen-9%—$0.02 / $0.04Llama 3.2 3B InstructunstatedOpen-24%—$0.02 / $0.02All 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 3.1 licencecommercial use with conditions
Benchmarks
all categoriesChat & Knowledge
Arena Text · overall1,293#158
Arena Text · Chinese1,215.2#156
Arena Text · English1,293.6#154
Arena Text · French1,261.3#126
Arena Text · German1,221.7#128
Arena Text · instruction following1,231.1#156
Arena Text · Japanese1,132.3#120
Arena Text · Korean1,140.5#129
Arena Text · non-English1,219.2#155
Arena Text · Russian1,233.6#154
Arena Text · Spanish1,253#135
MATH Level 536.7#25
Arena Text · overall · full cohort1,260.9#156
GPQA Diamond · Epoch44.2#109
OTIS Mock AIME 2024 20253.6#117
Writing & Language
Arena Text · creative writing1,257#156
EQ-Bench Creative Writing v3783.6#83
Arena Text · creative writing · full cohort1,231.7#153
FAQ
How much does Llama 3.1 70B Instruct cost?›
$0.40 per 1M input tokens and $0.40 per 1M output tokens (cheapest host: DeepInfra).
Is Llama 3.1 70B Instruct free to use?›
There is no free API; the cheapest paid price is $0.40 per 1M input tokens. The weights are open, so you can run it yourself.
How good is Llama 3.1 70B Instruct, and what is it best at?›
#158 of 196 in Chat & Knowledge, #156 of 195 in Writing & Language.
Is Llama 3.1 70B Instruct better than Nanbeige4 3B Thinking 2511?›
On EQ-Bench Creative Writing v3, Llama 3.1 70B Instruct scores 783.6 vs 841.2 — lower.
How can I use Llama 3.1 70B Instruct?›
Through the API of Amazon Bedrock, DeepInfra.
What is the context window of Llama 3.1 70B Instruct?›
131K tokens, with up to 16K output tokens.
Is Llama 3.1 70B Instruct open source?›
Yes, the weights are open (llama3.1 licence).
Can I run Llama 3.1 70B Instruct locally?›
The weights are open and can be self-hosted; hardware needs are not confirmed yet.
Can Llama 3.1 70B Instruct read images?›
No, it accepts text only.
What is Llama 3.1 70B Instruct?›
Llama 3.1 70B Instruct is an open-weights language model from Meta.
What are the best alternatives to Llama 3.1 70B Instruct?›
Gemma 4 31B (cheaper), GPT-6 Astra (better), Muse Glimmer 30B (open weights).