modelbenchmark.io

DeepSeek R1 Distill Llama 70B

DeepSeek · deepseek-r1-distill-llama-70b

Compare

DeepSeek reasoning model for multi-step analysis, math, coding, and tools

Specification

most-agreed values

Context
128K
Max output
4K
Released
2025-01-20
Knowledge cutoff
2025-01
Retires
2025-12-23
Open weights
no
Input
text
Output
text

Price

US dollars per million tokens · most-agreed

Input
$0.03
Output
$0.13

Quality

3 benchmarks · 1 source

Composite
66th
percentile of 344 scored models
Rank
118 / 344
±0.06 sd
Evidence
3 × 1
one source, or fewer than 3 benchmarks
Effort range
—
one configuration only
reasoning39th1/4 bench
math80th2/7 bench

Every score

one row per source and configuration — nothing averaged away

BenchmarkScoreConfigurationSourceRun
MATH level 589.9 ±0.7defaultEpoch AI2025-03-10
OTIS Mock AIME 2024-202551.4 ±6.1defaultEpoch AI2025-03-07
GPQA diamond55.7 ±3.0defaultEpoch AI2025-03-10

The leading figure for a benchmark is the median across its configurations, so one heroic high-effort run cannot set the number.

Available from 11 hosts

HostIn $/MOut $/MCache rdCache wrContextOutputRetires
helicone$0.03$0.13——128K4K—
deepinfra$0.20$0.60——131K131K—
nebius$0.25$0.75——128K128K—
alibaba-cn$0.287$0.861——33K16K—
nscale$0.375$0.375—————
ovhcloud · ovhcloud$0.67$0.67——131K131K—
vercel_ai_gateway · vercel-ai-gateway$0.75$0.99——131K131K—
novita$0.80$0.80——8K8K—
fireworks_ai$0.90$0.90——131K131K—
gradient_ai · gradient-ai$0.99$0.99——33K8K—
together_ai$2.00$2.00——131K131K2025-12-23

Price history

append-only observations · a listing writes a row only when its price moves

DateHostField$/M
2026-09-02openrouterInput$0.80
2026-09-06openrouterInput$0.80
2026-09-29openrouterInput$0.80
2026-09-02openrouterOutput$0.80
2026-09-06openrouterOutput$0.80
2026-09-29openrouterOutput$0.80
2026-09-15together_aiInput$2.00
2026-09-24together_aiInput$2.00
2026-09-15together_aiOutput$2.00
2026-09-24together_aiOutput$2.00

More from DeepSeek

most-hosted first