Ideal für: RAG (Retrieval-Augmented Generation)
Ranked on long-context accuracy, groundedness, and input-token price — RAG is input-token-heavy by design.
Aktualisiert April 2026. Top 3 diesen Monat: R1 0528, Hunyuan A13B Instruct, DeepSeek V3.
Podium
So werten wir
RAG workloads push enormous amounts of retrieved context through a model. The three things that matter: does it faithfully use what you retrieved (groundedness), does it degrade when the context is long (needle-in-a-haystack), and how much will a million input tokens cost you. Because RAG is input-heavy, the input price pillar gets a heavier weight than it does for agentic or generative workloads.
Our full methodology is published on the Methodik-Seite.
Säulen und Gewichte:
Full ranking
| Rang | Modell | Anbieter | Input $/1M | Output $/1M | Kontext |
|---|---|---|---|---|---|
| 1 | R1 0528 | DeepSeek | $0.50 | $2.15 | 163,840 |
| 2 | Hunyuan A13B Instruct | Tencent | $0.14 | $0.57 | 131,072 |
| 3 | DeepSeek V3 | DeepSeek | $0.32 | $0.89 | 163,840 |
| 4 | Qwen3.5 Plus 2026-02-15 | Qwen | $0.26 | $1.56 | 1,000,000 |
| 5 | Trinity Large Preview | Arcee AI | $0.00 | $0.00 | 131,000 |
| 6 | MiniMax M2.1 | MiniMax | $0.29 | $0.95 | 196,608 |
| 7 | Qwen3.5 397B A17B | Qwen | $0.39 | $2.34 | 262,144 |
| 8 | MiMo-V2-Flash | Xiaomi | $0.09 | $0.29 | 262,144 |
| 9 | MiniMax-01 | MiniMax | $0.20 | $1.10 | 1,000,192 |
| 10 | Llama 3.3 70B Instruct | Meta | $0.12 | $0.38 | 131,072 |
Field notes
A 1M+ token context window is usually overkill. Optimize retrieval quality first.
Prompt caching matters: pin the system prompt and retrieved context into the cache tier if available.
Use batch pricing for bulk backfills over your corpus.
FAQ
The questions teams ask before picking a model for rag (retrieval-augmented generation).
Get instant answers from our AI agent
