Informes diarios
Un informe por día, escrito a partir de los números que mide overfeed.news: precios de alquiler de GPU, precios de API de LLM y lo que publicó el mundo de la IA. Cada cifra de un informe sale de nuestros propios datos de ese día.
Mercado de GPU
The cheapest 24 GB VRAM GPU is the Tesla P40 at $0.1076/hr on vast.ai, while the MI300X leads the 192 GB tier at $2.39/hr on runpod; vast.ai dominates listings with 213 offers across 94 chips tracked.
Precios de API
OpenRouter tracks 468 models as of 2026-10-22, with 21 models currently free and 178 offering open weights; one new free model was added, one batch model was removed, and 12 models saw price changes including significant increases for several DeepSeek and Z.ai models.
Mercado de GPU
The cheapest 24 GB VRAM GPU is the RTX A5000 at $0.16/hr on RunPod, while the MI300X leads the 141+ GB tier at $2.39/hr on RunPod; vast.ai listings show the largest day-over-day swings, including a 96.7% drop for the RTX 4060 and an 81.8% jump for the Q RTX 6000.
Precios de API
On 2026-10-20, OpenRouter tracks 470 models with 20 free models and 177 open-weight models.
La semana en IA
Overfeed indexed 1,040 items across community, research, models and news sources during the week of October 12–18, with research papers from arXiv accounting for nearly half the volume and a single day, October 14, contributing 304 items.
Precios de API
On 2026-10-18, OpenRouter tracks 471 models with 19 free models and 180 open-weight models.
Mercado de GPU
The cheapest 24 GB VRAM GPU is the Tesla V100 at $0.0963/hr on vast.ai, while the MI300X leads the 141 GB+ tier at $2.39/hr on runpod; day-over-day moves show the RTX 4060 dropping 97.3% to $0.0822/hr and the RTX 2080 Ti rising 67.1% to $0.2028/hr.
Precios de API
On October 16, 2026, OpenRouter tracked 465 models with 18 free models and 177 open-weight models.
Mercado de GPU
The cheapest 24 GB VRAM tier is a Tesla V100 at $0.1341/hr on vast.ai, while the cheapest 141 GB tier is an MI300X at $2.39/hr on runpod; RTX 4060 surged 251.7% versus yesterday and RTX 5070 Ti rose 171% over the week.
Precios de API
OpenRouter tracks 463 models as of 2026-10-14 with 17 free models available, while DeepSeek V4 Pro 0423 more than doubled input price to $0.96 per million tokens and MoonshotAI Kimi Latest output price jumped to $24 per million tokens.
Mercado de GPU
The GPU rental market tracked 100 chip types across 222 listings from runpod and vast.ai on 2026-10-13, with the RTX PRO 4500 showing the largest single-day price increase at 102.7 percent and the GTX 1060 leading the weekly gain at 148.3 percent.
La semana en IA
Overfeed indexed 1,117 items across research, community, news, and models during the week of October 5–11, with research papers from arXiv accounting for 550 items and Hacker News discussions contributing 344 items.
Mercado de GPU
The GPU rental market tracked 98 chip types across 217 listings from runpod and vast.ai on 2026-10-11, with the cheapest 24 GB VRAM option at $0.1168/hr for a Tesla V100 on vast.ai and the highest-volume listing at 44 offers for B200 on vast.ai.
Precios de API
On 2026-10-10, OpenRouter tracks 458 models with 19 free models and 170 open-weight models.
Mercado de GPU
The cheapest 24 GB VRAM GPU is the RTX A5000 at $0.16/hr on runpod, while the MI300X at 192 GB VRAM leads the high end at $2.39/hr on runpod; vast.ai dominates day-over-day price moves with the RTX 5070 jumping 73.5% to $0.3289/hr and the RTX 4070 dropping 64.5% to $0.1348/hr.
Precios de API
OpenRouter tracks 467 models as of 2026-10-08, with 20 models currently free and two new Anthropic Claude Haiku 5.5 variants entering the paid tier; eleven existing models saw price adjustments while one model was removed.
Mercado de GPU
The cheapest GPU with at least 24 GB VRAM is the Tesla V100 at $0.0849/hr on vast.ai, while the H100 SXM jumped 50.4% to $2.69/hr on runpod compared to yesterday.
Precios de API
20 free models are available today on OpenRouter, with 464 models tracked and 177 offering open weights.
La semana en IA
Overfeed indexed 1,076 items this week, led by 550 research papers from arXiv and 331 Hacker News threads, with peak volume of 251 items on 29 September.
Precios de API
22 models are free on OpenRouter today, with DeepSeek Flash Latest now the cheapest paid option at $0.003 per million input tokens.