geekradar
Int
Explore GeekRadar

More to explore

Good gear All topics About us Say hello
GeekRadar / The comparison desk
COMPARE.
AI models
allenai

Llama-3.1-Tulu-3-70B

AI model

Context window
131,072 tokens
Total parameters
70.55 B
Open weights
Yes

Capability & preference

Human preference · Arena

Overall text preferences from head-to-head votes. Not factual accuracy or a task-success rate.

Llama-3.1-Tulu-3-70B 1255.7 Llama 3.1 Tulu 3 70B · 95% interval 1245.2–1266.1 · 2,846 votes · 2026-11-06

Recorded results

Only the same benchmark, version, workload, scenario and unit share a row. Missing results are not zero. System results include the full submitted hardware and software configuration.

Arena text preference / chinesehuman preference · 2026-11-06 · rating

Llama-3.1-Tulu-3-70B

1,249.38 rating

Confidence interval: 1,213.81 – 1,284.94

224 votes

Category
chinese
Evaluated configuration
Llama 3.1 Tulu 3 70B
Evaluation type
human preference
Arena text preference / codinghuman preference · 2026-11-06 · rating

Llama-3.1-Tulu-3-70B

1,235.15 rating

Confidence interval: 1,211.51 – 1,258.78

450 votes

Category
coding
Evaluated configuration
Llama 3.1 Tulu 3 70B
Evaluation type
human preference
Arena text preference / creative_writinghuman preference · 2026-11-06 · rating

Llama-3.1-Tulu-3-70B

1,230.96 rating

Confidence interval: 1,206.61 – 1,255.31

441 votes

Category
creative_writing
Evaluated configuration
Llama 3.1 Tulu 3 70B
Evaluation type
human preference
Arena text preference / englishhuman preference · 2026-11-06 · rating

Llama-3.1-Tulu-3-70B

1,272.68 rating

Confidence interval: 1,259.41 – 1,285.95

1629 votes

Category
english
Evaluated configuration
Llama 3.1 Tulu 3 70B
Evaluation type
human preference
Arena text preference / exclude_tieshuman preference · 2026-11-06 · rating

Llama-3.1-Tulu-3-70B

1,169.60 rating

Confidence interval: 1,152.20 – 1,187.00

1728 votes

Category
exclude_ties
Evaluated configuration
Llama 3.1 Tulu 3 70B
Evaluation type
human preference
Arena text preference / hard_promptshuman preference · 2026-11-06 · rating

Llama-3.1-Tulu-3-70B

1,219.78 rating

Confidence interval: 1,201.36 – 1,238.21

779 votes

Category
hard_prompts
Evaluated configuration
Llama 3.1 Tulu 3 70B
Evaluation type
human preference
Arena text preference / hard_prompts_englishhuman preference · 2026-11-06 · rating

Llama-3.1-Tulu-3-70B

1,236.37 rating

Confidence interval: 1,213.14 – 1,259.61

499 votes

Category
hard_prompts_english
Evaluated configuration
Llama 3.1 Tulu 3 70B
Evaluation type
human preference
Arena text preference / industry_business_and_management_and_financial_operationshuman preference · 2026-11-06 · rating

Llama-3.1-Tulu-3-70B

1,225.87 rating

Confidence interval: 1,193.81 – 1,257.93

288 votes

Category
industry_business_and_management_and_financial_operations
Evaluated configuration
Llama 3.1 Tulu 3 70B
Evaluation type
human preference
Arena text preference / industry_entertainment_and_sports_and_mediahuman preference · 2026-11-06 · rating

Llama-3.1-Tulu-3-70B

1,240.04 rating

Confidence interval: 1,217.84 – 1,262.24

551 votes

Category
industry_entertainment_and_sports_and_media
Evaluated configuration
Llama 3.1 Tulu 3 70B
Evaluation type
human preference
Arena text preference / industry_life_and_physical_and_social_sciencehuman preference · 2026-11-06 · rating

Llama-3.1-Tulu-3-70B

1,266.75 rating

Confidence interval: 1,242.56 – 1,290.93

520 votes

Category
industry_life_and_physical_and_social_science
Evaluated configuration
Llama 3.1 Tulu 3 70B
Evaluation type
human preference
Arena text preference / industry_mathematicalhuman preference · 2026-11-06 · rating

Llama-3.1-Tulu-3-70B

1,246.35 rating

Confidence interval: 1,218.55 – 1,274.15

349 votes

Category
industry_mathematical
Evaluated configuration
Llama 3.1 Tulu 3 70B
Evaluation type
human preference
Arena text preference / industry_software_and_it_serviceshuman preference · 2026-11-06 · rating

Llama-3.1-Tulu-3-70B

1,241.62 rating

Confidence interval: 1,221.93 – 1,261.32

711 votes

Category
industry_software_and_it_services
Evaluated configuration
Llama 3.1 Tulu 3 70B
Evaluation type
human preference
Arena text preference / industry_writing_and_literature_and_languagehuman preference · 2026-11-06 · rating

Llama-3.1-Tulu-3-70B

1,254.21 rating

Confidence interval: 1,235.05 – 1,273.37

758 votes

Category
industry_writing_and_literature_and_language
Evaluated configuration
Llama 3.1 Tulu 3 70B
Evaluation type
human preference
Arena text preference / instruction_followinghuman preference · 2026-11-06 · rating

Llama-3.1-Tulu-3-70B

1,232.42 rating

Confidence interval: 1,216.99 – 1,247.85

1170 votes

Category
instruction_following
Evaluated configuration
Llama 3.1 Tulu 3 70B
Evaluation type
human preference
Arena text preference / longer_queryhuman preference · 2026-11-06 · rating

Llama-3.1-Tulu-3-70B

1,223.97 rating

Confidence interval: 1,199.37 – 1,248.58

463 votes

Category
longer_query
Evaluated configuration
Llama 3.1 Tulu 3 70B
Evaluation type
human preference
Arena text preference / mathhuman preference · 2026-11-06 · rating

Llama-3.1-Tulu-3-70B

1,241.90 rating

Confidence interval: 1,217.44 – 1,266.35

397 votes

Category
math
Evaluated configuration
Llama 3.1 Tulu 3 70B
Evaluation type
human preference
Arena text preference / multi_turnhuman preference · 2026-11-06 · rating

Llama-3.1-Tulu-3-70B

1,251.50 rating

Confidence interval: 1,228.37 – 1,274.62

537 votes

Category
multi_turn
Evaluated configuration
Llama 3.1 Tulu 3 70B
Evaluation type
human preference
Arena text preference / non_englishhuman preference · 2026-11-06 · rating

Llama-3.1-Tulu-3-70B

1,235.43 rating

Confidence interval: 1,220.24 – 1,250.62

1217 votes

Category
non_english
Evaluated configuration
Llama 3.1 Tulu 3 70B
Evaluation type
human preference
Arena text preference / overallhuman preference · 2026-11-06 · rating

Llama-3.1-Tulu-3-70B

1,255.69 rating

Confidence interval: 1,245.24 – 1,266.14

2846 votes

Category
overall
Evaluated configuration
Llama 3.1 Tulu 3 70B
Evaluation type
human preference
Arena text preference / russianhuman preference · 2026-11-06 · rating

Llama-3.1-Tulu-3-70B

1,245.56 rating

Confidence interval: 1,218.59 – 1,272.53

366 votes

Category
russian
Evaluated configuration
Llama 3.1 Tulu 3 70B
Evaluation type
human preference
Arena text preference / chinesehuman preference · 2026-10-15 · rating

Llama-3.1-Tulu-3-70B

1,249.29 rating

Confidence interval: 1,213.70 – 1,284.87

224 votes

Category
chinese
Evaluated configuration
llama-3.1-tulu-3-70b
Evaluation type
human preference
Arena text preference / codinghuman preference · 2026-10-15 · rating

Llama-3.1-Tulu-3-70B

1,235.22 rating

Confidence interval: 1,211.57 – 1,258.87

450 votes

Category
coding
Evaluated configuration
llama-3.1-tulu-3-70b
Evaluation type
human preference
Arena text preference / creative_writinghuman preference · 2026-10-15 · rating

Llama-3.1-Tulu-3-70B

1,231.13 rating

Confidence interval: 1,206.76 – 1,255.49

441 votes

Category
creative_writing
Evaluated configuration
llama-3.1-tulu-3-70b
Evaluation type
human preference
Arena text preference / englishhuman preference · 2026-10-15 · rating

Llama-3.1-Tulu-3-70B

1,272.60 rating

Confidence interval: 1,259.33 – 1,285.88

1629 votes

Category
english
Evaluated configuration
llama-3.1-tulu-3-70b
Evaluation type
human preference
Arena text preference / exclude_tieshuman preference · 2026-10-15 · rating

Llama-3.1-Tulu-3-70B

1,169.66 rating

Confidence interval: 1,152.26 – 1,187.06

1728 votes

Category
exclude_ties
Evaluated configuration
llama-3.1-tulu-3-70b
Evaluation type
human preference
Arena text preference / hard_promptshuman preference · 2026-10-15 · rating

Llama-3.1-Tulu-3-70B

1,219.87 rating

Confidence interval: 1,201.43 – 1,238.30

779 votes

Category
hard_prompts
Evaluated configuration
llama-3.1-tulu-3-70b
Evaluation type
human preference
Arena text preference / hard_prompts_englishhuman preference · 2026-10-15 · rating

Llama-3.1-Tulu-3-70B

1,236.33 rating

Confidence interval: 1,213.07 – 1,259.58

499 votes

Category
hard_prompts_english
Evaluated configuration
llama-3.1-tulu-3-70b
Evaluation type
human preference
Arena text preference / industry_business_and_management_and_financial_operationshuman preference · 2026-10-15 · rating

Llama-3.1-Tulu-3-70B

1,226.01 rating

Confidence interval: 1,193.93 – 1,258.09

288 votes

Category
industry_business_and_management_and_financial_operations
Evaluated configuration
llama-3.1-tulu-3-70b
Evaluation type
human preference
Arena text preference / industry_entertainment_and_sports_and_mediahuman preference · 2026-10-15 · rating

Llama-3.1-Tulu-3-70B

1,240.08 rating

Confidence interval: 1,217.87 – 1,262.30

551 votes

Category
industry_entertainment_and_sports_and_media
Evaluated configuration
llama-3.1-tulu-3-70b
Evaluation type
human preference
Arena text preference / industry_life_and_physical_and_social_sciencehuman preference · 2026-10-15 · rating

Llama-3.1-Tulu-3-70B

1,266.76 rating

Confidence interval: 1,242.56 – 1,290.95

520 votes

Category
industry_life_and_physical_and_social_science
Evaluated configuration
llama-3.1-tulu-3-70b
Evaluation type
human preference
Arena text preference / industry_mathematicalhuman preference · 2026-10-15 · rating

Llama-3.1-Tulu-3-70B

1,246.22 rating

Confidence interval: 1,218.41 – 1,274.02

349 votes

Category
industry_mathematical
Evaluated configuration
llama-3.1-tulu-3-70b
Evaluation type
human preference
Arena text preference / industry_software_and_it_serviceshuman preference · 2026-10-15 · rating

Llama-3.1-Tulu-3-70B

1,241.73 rating

Confidence interval: 1,222.03 – 1,261.44

711 votes

Category
industry_software_and_it_services
Evaluated configuration
llama-3.1-tulu-3-70b
Evaluation type
human preference
Arena text preference / industry_writing_and_literature_and_languagehuman preference · 2026-10-15 · rating

Llama-3.1-Tulu-3-70B

1,254.29 rating

Confidence interval: 1,235.13 – 1,273.46

758 votes

Category
industry_writing_and_literature_and_language
Evaluated configuration
llama-3.1-tulu-3-70b
Evaluation type
human preference
Arena text preference / instruction_followinghuman preference · 2026-10-15 · rating

Llama-3.1-Tulu-3-70B

1,232.42 rating

Confidence interval: 1,216.99 – 1,247.86

1170 votes

Category
instruction_following
Evaluated configuration
llama-3.1-tulu-3-70b
Evaluation type
human preference
Arena text preference / longer_queryhuman preference · 2026-10-15 · rating

Llama-3.1-Tulu-3-70B

1,224.01 rating

Confidence interval: 1,199.38 – 1,248.63

463 votes

Category
longer_query
Evaluated configuration
llama-3.1-tulu-3-70b
Evaluation type
human preference
Arena text preference / mathhuman preference · 2026-10-15 · rating

Llama-3.1-Tulu-3-70B

1,241.85 rating

Confidence interval: 1,217.39 – 1,266.31

397 votes

Category
math
Evaluated configuration
llama-3.1-tulu-3-70b
Evaluation type
human preference
Arena text preference / multi_turnhuman preference · 2026-10-15 · rating

Llama-3.1-Tulu-3-70B

1,251.56 rating

Confidence interval: 1,228.43 – 1,274.70

537 votes

Category
multi_turn
Evaluated configuration
llama-3.1-tulu-3-70b
Evaluation type
human preference
Arena text preference / non_englishhuman preference · 2026-10-15 · rating

Llama-3.1-Tulu-3-70B

1,235.57 rating

Confidence interval: 1,220.38 – 1,250.76

1217 votes

Category
non_english
Evaluated configuration
llama-3.1-tulu-3-70b
Evaluation type
human preference
Arena text preference / overallhuman preference · 2026-10-15 · rating

Llama-3.1-Tulu-3-70B

1,255.69 rating

Confidence interval: 1,245.23 – 1,266.14

2846 votes

Category
overall
Evaluated configuration
llama-3.1-tulu-3-70b
Evaluation type
human preference
Arena text preference / russianhuman preference · 2026-10-15 · rating

Llama-3.1-Tulu-3-70B

1,245.61 rating

Confidence interval: 1,218.62 – 1,272.60

366 votes

Category
russian
Evaluated configuration
llama-3.1-tulu-3-70b
Evaluation type
human preference

Run it locally

Inference · memory estimate

Llama-3.1-Tulu-3-70B

70.55 B total parameters · llama · BF16

Published checkpoint tensor payload: 131.4 GiB. This includes stored copies and buffers; it is separate from the runtime estimates below.

Estimated accelerator memory41.0 GiB
Weights
36.1 GiB
KV cache
1.3 GiB
Runtime allowance
3.6 GiB

One user, full model in accelerator memory.

GPUs with memory headroom

Local AI systems with memory headroom

Shared-memory systems use a planning budget of 75% of installed memory, leaving 25% for the OS and other allocations. This is an assumption, not a measured GPU allocation. Firmware and driver limits may be lower and may require configuration. Memory matches do not verify a runtime, quantized model artifact or inference speed.

Includes FP16 KV cache, prompt plus generated tokens, quantization metadata and a runtime allowance. GPU matches reserve 10% of VRAM. Memory fit is not a speed or software-compatibility guarantee; quantized weights and runtime support must be available.

Capabilities

Context window
131,072 tokens
Open weights
Yes
Model license
llama3.1
Total parameters
70.55 B
Checkpoint tensor payload
131.42 GiB

Input modalities

  • text

Output modalities

  • text

Processing & platform

Architecture
llama