50 models7 providersprices scraped nightly — no estimates
meta · llama familyVerified May 27, 2026

Llama 3.1 8B Instruct

8B params · 131K context · llama-3 · released 2024-07 · 6 providers

Llama 3.1 8B Instruct is the workhorse of the small-model tier: 8 billion parameters, 131K context window, and per-token pricing that sits at $0.05–$0.10 across most providers — cheap enough to run at scale without budget tracking. It's the right choice for classification, simple summarization, content moderation, autocomplete-style code suggestions, and any workload where you'd otherwise reach for a custom-trained model. Quality is well below the 70B for complex reasoning or nuanced generation, but for the 80% of production LLM workloads that are "extract this field" or "categorize this email," 8B is more than enough — and lets you run 10x the volume for the same monthly bill.

Cheapest right now
DeepInfra$0.0600/1M out
sticky while scrolling · verified May 27, 2026
Cheapest now
$0.0600
DeepInfra · fp16 · $/1M out
90-day price Δ
cheapest-host trend
Fastest TTFT
50 ms
Groq · 800 tok/s
Providers
6
indexed providers
✓ verified May 27, 2026
#Provider$/1M in$/1M outTTFT p50tok/sQuant
1DeepInfraCHEAPEST$0.0600$0.0600160 ms420fp16go ↗
2GroqFASTEST$0.0500$0.080050 ms800fp8go ↗
3DeepInfra$0.0600$0.0600160 ms420fp16go ↗
4OpenRouter$0.0600$0.0900unknowngo ↗
5Fireworks AI$0.0700$0.0700150 ms500fp16go ↗
6Cerebras Inference$0.10$0.10100 ms1800fp16go ↗
7Hyperbolic$0.10$0.10220 ms280fp16go ↗

Prices verified May 27, 2026 by scraping 6 providers. Methodology · Raw data