OpenRouter provider table for meta-llama/llama-3.3-70b-instruct, fourth re-read 2026-10-17 (per-provider price, P50 latency, throughput, uptime)

Fetched later on 2026-10-17 than svc-src-openrouter-provider-table-2026-10-17-r3, same URL. Per-provider rows for Llama 3.3 70B Instruct, P50 over OpenRouter's own routed traffic. DeepInfra (Turbo) $0.100/$0.320, 1.71s latency, 6 tps, 99.92% uptime (was 0.55s / 16 tps / 99.94% on the r3 read). NovitaAI $0.135/$0.400, 1.05s, 41 tps, 99.29% (was 1.08s / 41 tps / 99.25%). AkashML $0.200/$0.520 (cache read $0.100), 0.59s, 40 tps, 99.85%. Parasail $0.220/$0.500 (cache read $0.110), 0.63s, 45 tps, 100.00%. Cloudflare $0.293/$2.253, 0.72s, 40 tps, 99.42%. SambaNova Turbo list $0.60/$1.20 struck through next to a 25% off badge rendering $0.450/$0.900, 2.58s, 162 tps, 99.86% (was 0.78s / 110 tps / 99.88%). Groq $0.590/$0.790 (cache read $0.295), 0.51s, 198 tps, 100.00% (was 0.53s / 201 tps). CoreWeave $0.710/$0.710 (cache read $0.710), 0.11s, 88 tps, 100.00% (was 0.16s / 86 tps). Google Vertex (US flag) $0.720/$0.720, 1.22s, 96 tps; a second Vertex row at the same price is 1.40s / 96 tps (r3's US row was 2.32s / 92 tps and the second row had no latency cell). Together $1.040/$1.040, 0.55s, 41 tps, 96.68% (was 0.71s / 39 tps / 96.30%). Page headline price is still the floor route $0.10/$0.32. Best-across-providers throughput widget reads 198 tok/s (Groq); best latency widget reads 0.11s (CoreWeave). Weighted average actually paid: $0.2506 in / $0.5767 out. List prices on every row are unchanged from the r3 read, including the SambaNova struck-through discount. These are OpenRouter-measured P50 numbers over its own traffic, not vendor claims and not the Artificial Analysis long-prompt medians.