The Known Good Updated 4 Sep 2026 Subscribe
DeepSeek: R1 Distill Llama 70B vs Google: Gemma 3 27B ·DeepSeek vs Google

DeepSeek: R1 Distill Llama 70B vs Google: Gemma 3 27B — head-to-head comparison

DeepSeek: R1 Distill Llama 70B profile Google: Gemma 3 27B profile All comparisons
DeepSeek: R1 Distill Llama 70B — DeepSeek
Quality
#41 of 65
63based on 4 of 7 evaluations
Known Good Index
Speed
#306 of 338
20.7
Output tokens per second
Price
#211 of 442
$0.80
Blended per 1M tokens
In $0.80 · Out $0.80
Cache hit price
unranked
—
per 1M tokens
Context
#467 of 476
8k
tokens (~13 pages)
Google: Gemma 3 27B — Google
Quality
#49 of 65
47based on 4 of 7 evaluations
Known Good Index
Speed
#311 of 338
19.0
Output tokens per second
Price
#73 of 442
$0.17
Blended per 1M tokens
In $0.08 · Out $0.45
Cache hit price
#104 of 324
$0.040
per 1M tokens (−50%)
Context
#373 of 476
131k
tokens (~200 pages)
Summary

On the Known Good Index, DeepSeek: R1 Distill Llama 70B scores 63 and Google: Gemma 3 27B scores 47 — a 16-point lead for DeepSeek: R1 Distill Llama 70B. They share 10 published evaluations, with DeepSeek: R1 Distill Llama 70B ahead on 8 and Google: Gemma 3 27B ahead on 2.

Blended pricing is $0.80 per 1M tokens for DeepSeek: R1 Distill Llama 70B and $0.17 for Google: Gemma 3 27B — Google: Gemma 3 27B is about 4.6× cheaper.

Throughput is 21 tokens per second for DeepSeek: R1 Distill Llama 70B and 19 for Google: Gemma 3 27B, with time to first token at 1.0s and 1.0s. DeepSeek: R1 Distill Llama 70B accepts 8k tokens of context against 131k.

DeepSeek: R1 Distill Llama 70B is open weights under Open weights (unrestricted); Google: Gemma 3 27B is open weights under Open weights (restricted use). 5 tracked hosts serve one or both, and 1 serve both.

Generated from tracked data on 4 Sep 2026 · how these comparisons are computed →

At a glance
MetricDeepSeek: R1 Distill Llama 70BGoogle: Gemma 3 27B
Known Good Index 63 based on 4 of 7 evaluations 47 based on 4 of 7 evaluations
Blended $/1M $0.80 $0.17
Output tok/s 21 19
Context window 8k 131k
Providers 1 5
Jump to section

Quality

Known Good Index and the capability indices, side by side

Index scores

Each index is min-max normalised across tracked models and scaled 0–100 from published evaluations. We do not run these evaluations.

MetricDeepSeek: R1 Distill Llama 70BGoogle: Gemma 3 27B
Known Good Index 63 47
Mathematics Index 71 49
Reasoning Index 59 40
Openness Index 60 60
The Known Good

Quality Breakdown

Every evaluation both models report a published score for

Evaluations both models report

Bar = DeepSeek: R1 Distill Llama 70B · marker = Google: Gemma 3 27B · scores ingested from Epoch AI (CC-BY 4.0), LiveBench (Apache-2.0)

GPQA Diamond56 / 48
Mock AIME 2024–2551 / 22
MATH Level 590 / 74
LiveBench54 / 50
LiveBench Coding52 / 40
LiveBench Data Analysis56 / 51
LiveBench Instruction Following70 / 75
LiveBench Language24 / 35
LiveBench Mathematics58 / 55
LiveBench Reasoning68 / 44

Right-hand figures read DeepSeek: R1 Distill Llama 70B / Google: Gemma 3 27B. DeepSeek: R1 Distill Llama 70B leads on 8 of 10, Google: Gemma 3 27B on 2.

The Known Good

Arena Elo

Human preference rating from blind pairwise votes

Arena ratings

Elo from blind pairwise votes, ingested from Arena. Higher is better.

MetricDeepSeek: R1 Distill Llama 70BGoogle: Gemma 3 27B
Text Arena — 1,358
Text Arena (style controlled) — 1,365
The Known Good

Price & Cost

Published API pricing by token type

Published pricing

USD per 1M tokens. Blended is a 3:1 input:output mix. Live from OpenRouter, cross-checked against models.dev. Lower is better.

MetricDeepSeek: R1 Distill Llama 70BGoogle: Gemma 3 27B
Input $/1M $0.80 $0.08
Output $/1M $0.80 $0.45
Blended $/1M $0.80 $0.17
Cache read $/1M — $0.040
The Known Good
We publish blended price rather than cost-per-task. Cost-per-task requires token counts from an evaluation suite we do not run. Blended price is computed from published rates and is fully reproducible — see the pricing methodology.

Context Window

Maximum input accepted and maximum output emitted

Context and output limits

Provider-declared limits, from metadata via models.dev and OpenRouter. Higher is better.

MetricDeepSeek: R1 Distill Llama 70BGoogle: Gemma 3 27B
Context window 8k 131k
Max output tokens 7k 118k
The Known Good

Speed & Latency

Mean throughput and time to first token across hosting providers, over successful measurements from the last 48 hours

Throughput and latency

Means across all tracked providers over successful measurements from the last 48 hours, derived from OpenRouter endpoint telemetry.

MetricDeepSeek: R1 Distill Llama 70BGoogle: Gemma 3 27B
Output tokens/s 21 19
Time to first token 1.0s 1.0s
The Known Good

Providers

Which hosts serve each model, and at what price and speed

5 hosts · 1 serve both
Full provider benchmarks →
Provider DeepSeek: R1 Distill Llama 70B blendedDeepSeek: R1 Distill Llama 70B tok/s Google: Gemma 3 27B blendedGoogle: Gemma 3 27B tok/s
Novita $0.80 17 $0.14 7
Phala — — $0.23 51
DeepInfra — — $0.10 16
Parasail — — $0.17 11
Nebius — — $0.15 7

Per-host pricing and telemetry from OpenRouter. A dash means that host does not serve that model.

The Known Good
Sources for this page
Pricing & providers — OpenRouter Metadata — models.dev Benchmarks — Epoch AI (CC-BY 4.0)Benchmarks — LiveBench (Apache-2.0)Elo — Arena Full attribution →