daily refresh 2026-11-10

gpu_price_history.csv

@@ -68,4 +68,4 @@ date,gpu,min_usd_per_gpu_hr,median_usd_per_gpu_hr,max_usd_per_gpu_hr,n_rows,n_pr
 2026-11-10,GB200,10.5,16.0,27.04,3,3
 2026-11-10,GB300,13.72,15.86,18.0,2,2
 2026-11-10,MI300X,3.45,4.725,6.0,2,2
-2026-11-10,MI355X,8.6,8.6,8.6,1,1
+2026-11-10,MI355X,5.99,7.295,8.6,2,2

gpu_price_rows_daily.csv

@@ -1474,5 +1474,7 @@ snapshot_date,provider,gpu,variant,gpus_per_unit,price_usd_per_gpu_hr,price_type
 2026-11-10,Verda,L40S,48GB,1,0.989,spot,none,not stated,per GPU-hour,$0.9890/h (spot),,,unknown,https://verda.com/pricing,2026-11-10,html_static,high,
 2026-11-10,Crusoe,MI300X,192GB,1,3.45,on_demand,none,not stated,per GPU-hour,$3.45/GPU-hr,,Page: contact sales to lock in guaranteed resources at lowest rates,unknown,https://www.crusoe.ai/cloud/pricing,2026-11-10,html_static,medium,Price column read by position (the page lists models and prices in separate blocks). Verify before relying on it.
 2026-11-10,Oracle Cloud (OCI),MI300X,,1,6.00,on_demand,none,global list price,per GPU-hour,$6 GPU Per Hour (B109485),,,unknown,https://www.oracle.com/cloud/price-list/,2026-11-10,official_api,high,"Oracle Cloud Price List API, lastUpdated 2026-11-02T14:49:49.342Z. Part: OCI - Compute - GPU - MI300X. GPU memory size is not stated in the price list. Capacity must be requested; not shown by the price list."
+2026-11-10,Daytona,MI355X,288GB,1,5.99,on_demand,,,per hour,"$5.99/h GPU only, on-demand",,,instant_self_serve,https://www.daytona.io/pricing,2026-11-10,manual,high,"GPU rate only; Daytona bills the minimum 1 vCPU, 1 GiB RAM and 1 GiB disk separately. Read by hand from the public pricing page. Recheck on each weekly pass."
 2026-11-10,Oracle Cloud (OCI),MI355X,,1,8.60,on_demand,none,global list price,per GPU-hour,$8.6 GPU Per Hour (B111758),,,unknown,https://www.oracle.com/cloud/price-list/,2026-11-10,official_api,high,"Oracle Cloud Price List API, lastUpdated 2026-11-02T14:49:49.342Z. Part: OCI- Compute - GPU - MI355X. GPU memory size is not stated in the price list. Capacity must be requested; not shown by the price list."
+2026-11-10,Daytona,MI355X,288GB,1,3.44,spot,,,per hour,"$3.44/h GPU only, interruptible",,,instant_self_serve,https://www.daytona.io/pricing,2026-11-10,manual,high,"GPU rate only, interruptible. Same page as the on-demand row."
 2026-11-10,Crusoe,MI355X,288GB,,,contact_sales,,,,Contact sales,,,contact_sales,https://www.crusoe.ai/cloud/pricing,2026-11-10,html_static,high,

gpu_prices.csv

@@ -208,5 +208,7 @@ Nebius,L40S,AMD CPU,1,0.90,preemptible,none,not stated,per GPU-hour,from $0.90**
 Verda,L40S,48GB,1,0.989,spot,none,not stated,per GPU-hour,$0.9890/h (spot),,,unknown,https://verda.com/pricing,2026-11-10,html_static,high,
 Crusoe,MI300X,192GB,1,3.45,on_demand,none,not stated,per GPU-hour,$3.45/GPU-hr,,Page: contact sales to lock in guaranteed resources at lowest rates,unknown,https://www.crusoe.ai/cloud/pricing,2026-11-10,html_static,medium,Price column read by position (the page lists models and prices in separate blocks). Verify before relying on it.
 Oracle Cloud (OCI),MI300X,,1,6.00,on_demand,none,global list price,per GPU-hour,$6 GPU Per Hour (B109485),,,unknown,https://www.oracle.com/cloud/price-list/,2026-11-10,official_api,high,"Oracle Cloud Price List API, lastUpdated 2026-11-02T14:49:49.342Z. Part: OCI - Compute - GPU - MI300X. GPU memory size is not stated in the price list. Capacity must be requested; not shown by the price list."
+Daytona,MI355X,288GB,1,5.99,on_demand,,,per hour,"$5.99/h GPU only, on-demand",,,instant_self_serve,https://www.daytona.io/pricing,2026-11-10,manual,high,"GPU rate only; Daytona bills the minimum 1 vCPU, 1 GiB RAM and 1 GiB disk separately. Read by hand from the public pricing page. Recheck on each weekly pass."
 Oracle Cloud (OCI),MI355X,,1,8.60,on_demand,none,global list price,per GPU-hour,$8.6 GPU Per Hour (B111758),,,unknown,https://www.oracle.com/cloud/price-list/,2026-11-10,official_api,high,"Oracle Cloud Price List API, lastUpdated 2026-11-02T14:49:49.342Z. Part: OCI- Compute - GPU - MI355X. GPU memory size is not stated in the price list. Capacity must be requested; not shown by the price list."
+Daytona,MI355X,288GB,1,3.44,spot,,,per hour,"$3.44/h GPU only, interruptible",,,instant_self_serve,https://www.daytona.io/pricing,2026-11-10,manual,high,"GPU rate only, interruptible. Same page as the on-demand row."
 Crusoe,MI355X,288GB,,,contact_sales,,,,Contact sales,,,contact_sales,https://www.crusoe.ai/cloud/pricing,2026-11-10,html_static,high,

models.csv

@@ -15,3 +15,6 @@ gemma-2-27b,Gemma 2 27B,27.2,dense,https://huggingface.co/google/gemma-2-27b,202
 deepseek-v3,DeepSeek V3,684.5,moe,https://huggingface.co/deepseek-ai/DeepSeek-V3,2026-11-04,"The published weights are already FP8 (about 1 byte per parameter), so the FP16 row is hypothetical. Mixture of experts: all experts must sit in memory."
 gpt-oss-120b,gpt-oss-120b,116.8,moe,https://huggingface.co/openai/gpt-oss-120b,2026-11-04,"The published weights are already about 4-bit (MXFP4), so the FP16 row is hypothetical. Mixture of experts: all experts must sit in memory."
 gpt-oss-20b,gpt-oss-20b,20.9,moe,https://huggingface.co/openai/gpt-oss-20b,2026-11-04,"The published weights are already about 4-bit (MXFP4), so the FP16 row is hypothetical. Mixture of experts: all experts must sit in memory."
+kimi-k3-1-pro,Kimi K3.1 Pro,2780.1,moe,https://huggingface.co/moonshotai/Kimi-K3.1-Pro,2026-11-10,"Parameter count from the Hugging Face safetensors total (2.78T, about 104B active per token). Mixture of experts: all experts must sit in memory, so the memory estimate uses total parameters. Published weights are about 4-bit (MXFP4), so the FP16 row is hypothetical."
+longcat-2-5,LongCat-2.5,1779.8,moe,https://huggingface.co/meituan-longcat/LongCat-2.5,2026-11-10,"Parameter count from the Hugging Face safetensors total (1.78T; Meituan describes it as 1.6T with about 48B active). Mixture of experts: all experts must sit in memory, so the memory estimate uses total parameters."
+qwen3-8-27b,Qwen3.8 27B,27.8,dense,https://huggingface.co/Qwen/Qwen3.8-27B,2026-11-10,