← variables.md

2026-08-19

57 change(s) in the model market

added to the catalog

• LiquidAI: LFM2.5-2.6B (free) (liquid/lfm-2.5-2.6b:free) added to the catalog: 128,000 ctx, $0/M in / $0/M out
• Z.ai: GLM 5.3 (z-ai/glm-5.3) added to the catalog: 1,048,576 ctx, $1.4/M in / $4.4/M out

model price changes

• ~deepseek/deepseek-v4-flash-latest: prompt price $0.079/M → $0.076/M (−2.7%)
• ~deepseek/deepseek-v4-flash-latest: completion price $0.157/M → $0.153/M (−2.7%)
• google/gemma-4-31b-it: prompt price $0.1/M → $0.09/M (−10.0%)
• minimax/minimax-m3:batch: prompt price $0.15/M → $0.3/M (+100.0%)
• minimax/minimax-m3:batch: completion price $0.6/M → $1.2/M (+100.0%)
• moonshotai/kimi-k2.5: prompt price $0.57/M → $0.45/M (−21.1%)
• moonshotai/kimi-k2.5: completion price $2.85/M → $2.25/M (−21.1%)
• moonshotai/kimi-k2.7-code:batch: prompt price $0.475/M → $0.95/M (+100.0%)
• moonshotai/kimi-k2.7-code:batch: completion price $2/M → $4/M (+100.0%)
• nvidia/nemotron-3-ultra-550b-a55b:batch: prompt price $0.3/M → $0.6/M (+100.0%)
• nvidia/nemotron-3-ultra-550b-a55b:batch: completion price $1.8/M → $3.6/M (+100.0%)
• qwen/qwen3-next-80b-a3b-instruct: prompt price $0.1/M → $0.09/M (−10.0%)
• qwen/qwen3.5-122b-a10b: prompt price $0.29/M → $0.26/M (−10.3%)
• qwen/qwen3.5-122b-a10b: completion price $2.4/M → $2.08/M (−13.3%)
• qwen/qwen3.6-27b: prompt price $0.6/M → $0.3/M (−50.0%)
• qwen/qwen3.6-27b: completion price $3.6/M → $2/M (−44.4%)
• thinkingmachines/inkling:batch: prompt price $0.5/M → $1/M (+100.0%)
• thinkingmachines/inkling:batch: completion price $2.02/M → $4.05/M (+100.0%)
• z-ai/glm-5.2: prompt price $0.5/M → $0.966/M (+93.2%)
• z-ai/glm-5.2: completion price $3.15/M → $3.04/M (−3.6%)
• z-ai/glm-5.2:batch: prompt price $0.7/M → $1.4/M (+100.0%)
• z-ai/glm-5.2:batch: completion price $2.2/M → $4.4/M (+100.0%)

model context changes

• z-ai/glm-5.2:free: context length 128,000 → 256,000

quantization changes

• DeepSeek: deepseek/deepseek-v4-flash-0731 quantization fp8 → unknown

provider price changes

• Mancer 2: deepseek/deepseek-v4-flash prompt price $0.14/M → $0.145/M (+3.6%)
• Decart: deepseek/deepseek-v4-flash-0731 prompt price $0.079/M → $0.076/M (−3.4%)
• Decart: deepseek/deepseek-v4-flash-0731 completion price $0.158/M → $0.153/M (−3.4%)
• GMICloud: deepseek/deepseek-v4-flash-0731 prompt price $0.084/M → $0.112/M (+33.3%)
• GMICloud: deepseek/deepseek-v4-flash-0731 completion price $0.168/M → $0.224/M (+33.3%)
• Mancer 2: deepseek/deepseek-v4-flash-0731 prompt price $0.14/M → $0.145/M (+3.6%)
• Alibaba: deepseek/deepseek-v4-pro-0813 prompt price $1.45/M → $1.32/M (−9.1%)
• Alibaba: deepseek/deepseek-v4-pro-0813 completion price $4.36/M → $3.96/M (−9.1%)
• GMICloud: deepseek/deepseek-v4-pro-0813 prompt price $1.22/M → $1.19/M (−2.5%)
• GMICloud: deepseek/deepseek-v4-pro-0813 completion price $2.44/M → $3.56/M (+46.3%)
• Phala: meta/muse-glimmer-30b completion price $1.2/M → $1.1/M (−8.3%)
• Chutes: qwen/qwen3.8-27b prompt price $0.4/M → $0.45/M (+12.5%)
• Chutes: qwen/qwen3.8-27b completion price $3/M → $3.2/M (+6.7%)
• Io Net: qwen/qwen3.8-27b prompt price $0.44/M → $0.48/M (+9.1%)
• Io Net: qwen/qwen3.8-27b completion price $3.15/M → $3.4/M (+7.9%)
• Ambient: z-ai/glm-5.2 prompt price $1.05/M → $0.6/M (−42.9%)
• Ambient: z-ai/glm-5.2 completion price $4.4/M → $2/M (−54.5%)
• Phala: z-ai/glm-5.2 prompt price $1.13/M → $1.26/M (+11.1%)
• StreamLake: z-ai/glm-5.2 prompt price $0.56/M → $0.734/M (+31.0%)
• StreamLake: z-ai/glm-5.2 completion price $1.76/M → $2.31/M (+31.0%)

provider context changes

• Io Net: qwen/qwen3.8-27b context length 262,100 → 65,500

providers added

• OpenInference now serves deepseek/deepseek-v4-flash-0731 (fp4, 262,144 ctx, $0.08/M in / $0.18/M out)
• StreamLake now serves deepseek/deepseek-v4-pro-0813 (unknown, 1,024,000 ctx, $1.32/M in / $3.96/M out)
• Parasail now serves meta/muse-glimmer-30b (bf16, 131,072 ctx, $0.3/M in / $1.1/M out)

providers removed

• Phala no longer serves deepseek/deepseek-v4-pro-0813 (was unknown)
• Morph no longer serves google/gemma-4-31b-it (was fp4)
• Nebius no longer serves minimax/minimax-m2.5 (was fp4)
• DigitalOcean no longer serves qwen/qwen3-coder-30b-a3b-instruct (was unknown)
• Chutes no longer serves qwen/qwen3.5-397b-a17b (was fp8)
• Io Net no longer serves qwen/qwen3.6-27b (was fp8)
• Morph no longer serves qwen/qwen3.6-27b (was fp4)

openrouter catalog snapshots 2026-08-18 → 2026-08-19 · schema 1 · generated 2026-08-19T05:47:12.659Z