provider: Add Nscale

providers/nscale/models/Qwen/Qwen2.5-Coder-32B-Instruct.toml

@@ -0,0 +1,23 @@
+name = "Qwen 2.5 Coder 32B Instruct"
+description = "Large coding assistant for complex programming, refactoring, and code understanding"
+attachment = false
+reasoning = false
+tool_call = true
+structured_output = true
+temperature = true
+release_date = "2024-11-12"
+last_updated = "2024-11-12"
+open_weights = true
+
+[cost]
+input = 0.06
+output = 0.20
+
+[limit]
+context = 131_072
+input = 128_000
+output = 8_192
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/nscale/models/Qwen/Qwen2.5-Coder-3B-Instruct.toml

@@ -0,0 +1,23 @@
+name = "Qwen 2.5 Coder 3B Instruct"
+description = "Compact coding assistant for code completion and generation"
+attachment = false
+reasoning = false
+tool_call = true
+structured_output = true
+temperature = true
+release_date = "2024-11-12"
+last_updated = "2024-11-12"
+open_weights = true
+
+[cost]
+input = 0.01
+output = 0.03
+
+[limit]
+context = 32_768
+input = 30_000
+output = 8_192
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/nscale/models/Qwen/Qwen2.5-Coder-7B-Instruct.toml

@@ -0,0 +1,23 @@
+name = "Qwen 2.5 Coder 7B Instruct"
+description = "Mid-size coding assistant for code completion, generation, and understanding"
+attachment = false
+reasoning = false
+tool_call = true
+structured_output = true
+temperature = true
+release_date = "2024-11-12"
+last_updated = "2024-11-12"
+open_weights = true
+
+[cost]
+input = 0.01
+output = 0.03
+
+[limit]
+context = 32_768
+input = 30_000
+output = 8_192
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/nscale/models/Qwen/Qwen3-235B-A22B-Instruct-2507.toml

@@ -0,0 +1,16 @@
+base_model = "alibaba/qwen3-235b-a22b"
+
+name = "Qwen3 235B A22B Instruct 2507"
+description = "Large Mixture-of-Experts instruction model for general-purpose chat and tool use"
+release_date = "2025-07-18"
+last_updated = "2025-07-18"
+reasoning = false
+structured_output = true
+
+[cost]
+input = 0.20
+output = 0.60
+
+[limit]
+input = 128_000
+output = 8_192

providers/nscale/models/Qwen/Qwen3-235B-A22B.toml

@@ -0,0 +1,7 @@
+base_model = "alibaba/qwen3-235b-a22b"
+
+reasoning_options = []
+
+[cost]
+input = 0.20
+output = 0.60

providers/nscale/models/Qwen/Qwen3-4B-Instruct-2507.toml

@@ -0,0 +1,23 @@
+name = "Qwen3 4B Instruct 2507"
+description = "Compact instruction model for lightweight chat and generation tasks"
+attachment = false
+reasoning = false
+tool_call = true
+structured_output = true
+temperature = true
+release_date = "2025-07-18"
+last_updated = "2025-07-18"
+open_weights = true
+
+[cost]
+input = 0.01
+output = 0.03
+
+[limit]
+context = 131_072
+input = 128_000
+output = 8_192
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/nscale/models/Qwen/Qwen3-4B-Thinking-2507.toml

@@ -0,0 +1,24 @@
+name = "Qwen3 4B Thinking 2507"
+description = "Compact reasoning model with multi-step thinking capabilities"
+attachment = false
+reasoning = true
+reasoning_options = []
+tool_call = true
+structured_output = true
+temperature = true
+release_date = "2025-07-18"
+last_updated = "2025-07-18"
+open_weights = true
+
+[cost]
+input = 0.01
+output = 0.03
+
+[limit]
+context = 131_072
+input = 128_000
+output = 8_192
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/nscale/models/deepseek-ai/DeepSeek-R1-Distill-Llama-8B.toml

@@ -0,0 +1,27 @@
+name = "DeepSeek R1 Distill Llama 8B"
+description = "Compact distilled reasoning model based on Llama architecture"
+attachment = false
+reasoning = true
+reasoning_options = []
+tool_call = false
+structured_output = false
+temperature = true
+release_date = "2025-01-20"
+last_updated = "2025-01-20"
+open_weights = true
+
+[cost]
+input = 0.05
+output = 0.05
+
+[limit]
+context = 131_072
+input = 128_000
+output = 8_192
+
+[modalities]
+input = ["text"]
+output = ["text"]
+
+[interleaved]
+field = "reasoning_content"

providers/nscale/models/deepseek-ai/DeepSeek-R1-Distill-Qwen-14B.toml

@@ -0,0 +1,27 @@
+name = "DeepSeek R1 Distill Qwen 14B"
+description = "Mid-size distilled reasoning model based on Qwen architecture"
+attachment = false
+reasoning = true
+reasoning_options = []
+tool_call = false
+structured_output = false
+temperature = true
+release_date = "2025-01-20"
+last_updated = "2025-01-20"
+open_weights = true
+
+[cost]
+input = 0.20
+output = 0.20
+
+[limit]
+context = 131_072
+input = 128_000
+output = 8_192
+
+[modalities]
+input = ["text"]
+output = ["text"]
+
+[interleaved]
+field = "reasoning_content"

providers/nscale/models/deepseek-ai/DeepSeek-R1-Distill-Qwen-7B.toml

@@ -0,0 +1,27 @@
+name = "DeepSeek R1 Distill Qwen 7B"
+description = "Compact distilled reasoning model based on Qwen architecture optimized for math"
+attachment = false
+reasoning = true
+reasoning_options = []
+tool_call = false
+structured_output = false
+temperature = true
+release_date = "2025-01-20"
+last_updated = "2025-01-20"
+open_weights = true
+
+[cost]
+input = 0.15
+output = 0.15
+
+[limit]
+context = 131_072
+input = 128_000
+output = 8_192
+
+[modalities]
+input = ["text"]
+output = ["text"]
+
+[interleaved]
+field = "reasoning_content"

providers/nscale/models/meta-llama/Llama-3.1-8B-Instruct.toml

@@ -0,0 +1,24 @@
+name = "Llama 3.1 8B Instruct"
+description = "Compact open Llama model for general chat and instruction following"
+attachment = false
+reasoning = false
+tool_call = true
+structured_output = true
+temperature = true
+knowledge = "2023-12"
+release_date = "2024-07-23"
+last_updated = "2024-07-23"
+open_weights = true
+
+[cost]
+input = 0.06
+output = 0.06
+
+[limit]
+context = 131_072
+input = 128_000
+output = 8_192
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/nscale/models/meta-llama/Llama-3.2-11B-Vision-Instruct.toml

@@ -0,0 +1,24 @@
+name = "Llama 3.2 11B Vision Instruct"
+description = "Multimodal vision-language model for image understanding and text generation"
+attachment = true
+reasoning = false
+tool_call = true
+structured_output = true
+temperature = true
+knowledge = "2023-12"
+release_date = "2024-09-25"
+last_updated = "2024-09-25"
+open_weights = true
+
+[cost]
+input = 0.06
+output = 0.06
+
+[limit]
+context = 131_072
+input = 128_000
+output = 8_192
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]

providers/nscale/models/meta-llama/Llama-4-Scout-17B-16E-Instruct.toml

@@ -0,0 +1,12 @@
+base_model = "meta/llama-4-scout-17b-instruct"
+
+structured_output = true
+
+[cost]
+input = 0.09
+output = 0.29
+
+[limit]
+context = 131_072
+input = 128_000
+output = 8_192

providers/nscale/models/mistralai/Devstral-Small-2505.toml

@@ -0,0 +1,24 @@
+name = "Devstral Small 2505"
+description = "Mistral coding agent model for repository tasks and software engineering workflows"
+attachment = false
+reasoning = false
+tool_call = true
+structured_output = true
+temperature = true
+knowledge = "2025-05"
+release_date = "2025-05-21"
+last_updated = "2025-05-21"
+open_weights = true
+
+[cost]
+input = 0.10
+output = 0.30
+
+[limit]
+context = 128_000
+input = 124_000
+output = 8_192
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/nscale/models/mistralai/Mixtral-8x22B-Instruct-v0.1.toml

@@ -0,0 +1,24 @@
+name = "Mixtral 8x22B Instruct v0.1"
+description = "Large Mixture-of-Experts model for general-purpose instruction following"
+attachment = false
+reasoning = false
+tool_call = true
+structured_output = true
+temperature = true
+knowledge = "2024-04"
+release_date = "2024-04-17"
+last_updated = "2024-04-17"
+open_weights = true
+
+[cost]
+input = 1.20
+output = 1.20
+
+[limit]
+context = 65_536
+input = 63_000
+output = 4_096
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/nscale/models/openai/gpt-oss-120b.toml

@@ -0,0 +1,9 @@
+# reasoning_effort "low"|"medium"|"high" supported via OpenAI-compatible chat completions
+# https://docs.nscale.com/api-reference/inference/create-chat-completion (accessed 2026-07-13)
+base_model = "openai/gpt-oss-120b"
+
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
+
+[cost]
+input = 0.10
+output = 0.40

providers/nscale/models/openai/gpt-oss-20b.toml

@@ -0,0 +1,9 @@
+# reasoning_effort "low"|"medium"|"high" supported via OpenAI-compatible chat completions
+# https://docs.nscale.com/api-reference/inference/create-chat-completion (accessed 2026-07-13)
+base_model = "openai/gpt-oss-20b"
+
+reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
+
+[cost]
+input = 0.05
+output = 0.20

providers/nscale/provider.toml

@@ -0,0 +1,8 @@
+# Pricing and model details: https://nscale.com/serverless (accessed 2026-07-13)
+# API docs: https://docs.nscale.com/docs/use-cases/chat (accessed 2026-07-13)
+# Base URL and auth: https://docs.nscale.com/docs/use-cases/chat (accessed 2026-07-13)
+name = "Nscale"
+env = ["NSCALE_API_KEY"]
+npm = "@ai-sdk/openai-compatible"
+api = "https://inference.api.nscale.com/v1"
+doc = "https://docs.nscale.com/docs/ai-services/models"