← variables.md

2026-09-03

54 change(s) in the model market

added to the catalog

• Google: Gemini 3.8 Flash (google/gemini-3.8-flash) added to the catalog: 1,048,576 ctx, $0.75/M in / $3.75/M out
• Google: Gemini 3.8 Flash (batch) (google/gemini-3.8-flash:batch) added to the catalog: 1,048,576 ctx, $0.375/M in / $1.88/M out
• Meta: Muse Spark 1.3 (meta/muse-spark-1.3) added to the catalog: 1,048,576 ctx, $1.25/M in / $4.25/M out
• Meta: Muse Spark 1.3 Contributor (meta/muse-spark-1.3-contributor) added to the catalog: 1,048,576 ctx, $0.1/M in / $0.2/M out

removed from the catalog

• NVIDIA: Nemotron 3 Ultra (batch) (nvidia/nemotron-3-ultra-550b-a55b:batch) removed from the catalog

model price changes

• ~z-ai/glm-latest: prompt price $1.17/M → $1.11/M (−5.5%)
• ~z-ai/glm-latest: completion price $3.96/M → $3.48/M (−12.2%)
• google/gemini-3.7-flash:batch: prompt price $0.188/M → $0.375/M (+100.0%)
• google/gemini-3.7-flash:batch: completion price $0.938/M → $1.88/M (+100.0%)
• meta-llama/llama-3.3-70b-instruct: prompt price $0.71/M → $0.1/M (−85.9%)
• meta-llama/llama-3.3-70b-instruct: completion price $0.71/M → $0.32/M (−54.9%)
• meta/muse-glimmer-30b: completion price $1.2/M → $1.1/M (−8.3%)
• nvidia/nemotron-3-ultra-550b-a55b: prompt price $0.625/M → $0.6/M (−4.0%)
• nvidia/nemotron-3-ultra-550b-a55b: completion price $3.13/M → $2.4/M (−23.2%)
• qwen/qwen3.5-397b-a17b: prompt price $0.39/M → $0.55/M (+41.0%)
• qwen/qwen3.5-397b-a17b: completion price $2.34/M → $3.5/M (+49.6%)
• z-ai/glm-4.6: prompt price $0.43/M → $0.55/M (+27.9%)
• z-ai/glm-4.6: completion price $1.75/M → $2.2/M (+25.7%)

quantization changes

• AkashML: openai/gpt-oss-20b quantization bf16 → fp4

provider price changes

• Mancer 2: deepseek/deepseek-v4-flash prompt price $0.185/M → $0.175/M (−5.4%)
• Mancer 2: deepseek/deepseek-v4-flash-0731 prompt price $0.185/M → $0.175/M (−5.4%)
• Wafer: deepseek/deepseek-v4-flash-0731 prompt price $0.2/M → $0.1/M (−50.0%)
• Decart: moonshotai/kimi-k2.6 prompt price $0.537/M → $0.519/M (−3.3%)
• Decart: moonshotai/kimi-k2.6 completion price $2.26/M → $2.45/M (+8.3%)
• Morph: moonshotai/kimi-k3 prompt price $2.8/M → $2.6/M (−7.1%)
• Wafer: moonshotai/kimi-k3 completion price $15/M → $12.8/M (−15.0%)
• Mancer 2: openai/gpt-oss-120b prompt price $0.065/M → $0.06/M (−7.7%)
• AkashML: qwen/qwen3.8-27b prompt price $0.35/M → $0.32/M (−8.6%)
• AkashML: qwen/qwen3.8-27b completion price $2.55/M → $2.5/M (−2.0%)
• Parasail: qwen/qwen3.8-27b prompt price $0.35/M → $0.25/M (−28.6%)
• Parasail: qwen/qwen3.8-27b completion price $3.2/M → $2.2/M (−31.2%)
• Reka: qwen/qwen3.8-27b prompt price $0.35/M → $0.25/M (−28.6%)
• AtlasCloud: z-ai/glm-5.2 prompt price $1.26/M → $0.938/M (−25.6%)
• AtlasCloud: z-ai/glm-5.2 completion price $3.96/M → $2.95/M (−25.6%)
• Novita: z-ai/glm-5.2 prompt price $0.742/M → $0.559/M (−24.7%)
• Novita: z-ai/glm-5.2 completion price $2.33/M → $1.76/M (−24.7%)
• StreamLake: z-ai/glm-5.2 prompt price $0.693/M → $0.56/M (−19.2%)
• StreamLake: z-ai/glm-5.2 completion price $2.18/M → $1.76/M (−19.2%)
• Decart: z-ai/glm-5.3 prompt price $1.4/M → $1.11/M (−21.0%)
• Decart: z-ai/glm-5.3 completion price $4.4/M → $3.48/M (−21.0%)
• Inceptron: z-ai/glm-5.3 prompt price $1.4/M → $1.25/M (−10.7%)
• Reka: z-ai/glm-5.3 prompt price $1.4/M → $1.15/M (−17.9%)
• Reka: z-ai/glm-5.3 completion price $4.4/M → $3.5/M (−20.5%)
• Wafer: z-ai/glm-5.3 prompt price $1.4/M → $1.19/M (−15.0%)
• Wafer: z-ai/glm-5.3-flash prompt price $0.15/M → $0.1/M (−33.3%)
• Wafer: z-ai/glm-5.3-flash completion price $0.5/M → $0.35/M (−30.0%)

providers added

• DeepInfra now serves deepseek/deepseek-v4-flash-vision-exp (fp8, 1,048,576 ctx, $0.216/M in / $0.647/M out)
• DeepInfra now serves ibm-granite/granite-4.2-8b (bf16, 131,072 ctx, $0.06/M in / $0.25/M out)
• BaseTen now serves z-ai/glm-5.3 (fp8, 1,048,576 ctx, $2.1/M in / $6.6/M out)
• Sail Research now serves z-ai/glm-5.3-flash (fp8, 1,048,576 ctx, $0.15/M in / $0.5/M out)

providers removed

• Sail Research no longer serves deepseek/deepseek-v4-flash-0731 (was fp4)
• Phala no longer serves google/gemma-4-31b-it (was unknown)
• Mistral no longer serves mistralai/mistral-large (was unknown)
• AtlasCloud no longer serves z-ai/glm-5 (was fp8)

openrouter catalog snapshots 2026-09-02 → 2026-09-03 · schema 1 · generated 2026-09-03T09:51:23.387Z