ASCA06
MODEL DECISION PAGE

AI MODEL ATLAS · DECISION PAGE

Long-context LLMs for a 128K + 4K workload

Models whose listed context can fit 128,000 input plus 4,000 output tokens.

OBJECT IDasca:openrouter-model-page-long-context-llm@2026-08-27.v1
FETCHED
STATUSREPRODUCIBLE SNAPSHOT
QUESTION · METHOD

Long-context LLMs for a 128K + 4K workload

Keep context >= 132,000 and positive prompt/completion prices; derived_cost_usd = 128,000 × prompt + 4,000 × completion; request price is reported separately and excluded.

OBSERVED MODELS · 10 SHOWN

Comparable fields from one dated catalog snapshot

RANK ≠ QUALITY
Rank / orderModelInput / 1MOutput / 1MRequestContextToolsStructuredDerived cost / basis
73Ling-3.0-flash$0.021$0.063$0.00262.14KYesNo$0.0029 · $ per 128K input + 4K output text tokens
140Nex AGI: Nex-N2-Mini$0.025$0.10$0.00262.14KYesYes$0.0036 · $ per 128K input + 4K output text tokens
202OpenAI: GPT-5 Nano (batch)$0.025$0.20$0.00400KYesYes$0.004 · $ per 128K input + 4K output text tokens
388DeepSeek V4 Flash Latest$0.03$0.075$0.001.31MYesYes$0.0041 · $ per 128K input + 4K output text tokens
23Upstage: Solar Pro 4$0.03$0.12$0.00524.29KYesYes$0.0043 · $ per 128K input + 4K output text tokens
50Qwen: Qwen3.7 Flash$0.03$0.13$0.001MYesNo$0.0044 · $ per 128K input + 4K output text tokens
96Qwen: Qwen3 30B A3B Instruct 2507$0.0482$0.1931$0.00262.14KYesYes$0.0069 · $ per 128K input + 4K output text tokens
219Google: Gemini 2.5 Flash Lite (batch)$0.05$0.20$0.001.05MYesYes$0.0072 · $ per 128K input + 4K output text tokens
177NVIDIA: Nemotron 3 Nano 30B A3B$0.05$0.20$0.00262.14KYesYes$0.0072 · $ per 128K input + 4K output text tokens
373OpenAI: GPT-4.1 Nano (batch)$0.05$0.20$0.001.05MYesYes$0.0072 · $ per 128K input + 4K output text tokens

Prices are listed USD per token converted to USD per 1M tokens. Request fee is a separate listed field; it is not included in derived text-token cost. A free listing means zero listed token price in this snapshot, not unlimited access.

RANGE · BIAS · LIMIT · CONFIDENCE
  • Observed range: 275 matching catalog records; 10 shown.
  • Bias: Context metadata and price selection bias.
  • Limitation: A listed context limit does not prove usable quality, latency, or successful long-document completion.
  • Confidence: high for the recorded response and stated transformation.

DOWNLOAD · VERIFY · CITE

Keep the snapshot and object ID with the decision.

Object JSONDated source JSONNormalized CSVSource API ↗

Parent model atlas

“Keep context >= 132,000 and positive prompt/completion prices; derived_cost_usd = 128,000 × prompt + 4,000 × completion; request price is reported separately and excluded.”— asca:openrouter-model-page-long-context-llm@2026-08-27.v1
RELATED DECISION PAGES