Skip to content
OSMEXO

Hermes 3 70B Instruct

nous-research/hermes-3-llama-3.1-70b
  • Nous Research
  • Family · Hermes 3
  • Released 15 Aug 2024
  • Open weights
  • Text → Text
  • JSON
Not on OSMEXO yet
Spec & list price
Context128K
Max output16K
Input / 1M$0.70
Output / 1M$0.701× input
Scores · standing
Intelligence6.0#393/490
Usage
Tokens / day131.9M▼ 8.4% 7d
Usage rank#293
Requests / day36.9K
Per request · 30 days
Avg input2.6K0.2× median
Avg output1620.2× median

Consumption

83.3M166.7M250.0M3 Oct17 Oct1 Nov
Peak 218.6M 8 OctLow 101.6M 1 NovRank today #293

Benchmarks

bar = share of the top score · Benchmark data: Artificial Analysis
MMLU-Pro#211 of 273
57.1%
MATH-500#135 of 167
53.8%
GPQA Diamond#363 of 445
40.1%
LiveCodeBench#205 of 269
18.8%
Intelligence index#393 of 490
6.0
Humanity's Last Exam#405 of 465
4.0%
AIME#138 of 160
2.3%
Speed · median across providers, latest day
Output speed#295/33430 tok/s2.2× slower than median
First token#1370.88 s1.2× quicker than median

#18 cheapest of 31 with a similar score

Value
Intelligence Index
681012$0.2$0.5$1Hermes 3 70B Instruct
Blended price per 1M tokens (log)

Technical details

Tokenizer
Llama3
Knowledge cutoff
2023-12-31
License
llama3
Parameters
70.6B
Moderation
none
Downloads · likes
2.1K · 145
Request parameters 12
  • structured_outputs
  • temperature
  • top_p
  • top_k
  • min_p
  • max_tokens
  • stop
  • seed
  • frequency_penalty
  • logit_bias
+2 more
  • presence_penalty
  • repetition_penalty

When to use

Best for

  • Low-cost self-hosted chat
  • Steerable personas and role-play
  • Structured JSON output from text
  • Simple summarization and rewriting

Think twice for

  • Hard math and science reasoning
  • Competitive coding tasks
  • Image input
  • Agents that need native tool calling

Same family

Hermes 3

Alternatives

other vendors · similar capability

Want Hermes 3 70B Instruct on OSMEXO?

Sign in to request it. Requests are reviewed; a request does not guarantee availability.