Aion-RP 1.0 (8B) vs Qwen2.5-VL 72B Instruct vs Saba
Too close to call on our weighted score (Saba 46, Qwen2.5-VL 72B Instruct 44, Aion-RP 1.0 (8B) 24). The right pick depends on what you value most.
Aion Labs
Aion-RP 1.0 (8B)
24/100- ECI—
- Price$0.80 / $1.60
- Context33K
Alibaba (Qwen)Qwen2.5-VL 72B Instruct
44/100- ECI—
- Price$0.905 / $1.00
- Context131K
Mistral AISaba
46/100- ECI—
- Price$0.20 / $0.60
- Context33K
Too close to call
It is close. Our weighted score puts them within 2 points (Saba 46/100, Qwen2.5-VL 72B Instruct 44/100, Aion-RP 1.0 (8B) 24/100), so choose by what matters most for your work: Saba on price and Qwen2.5-VL 72B Instruct for long inputs. The score weighs price 50%, inputs & features 30%, context window 20%. None of the shared benchmarks cover every model here yet, so capability is left out of this verdict.
- CapabilityNot enough dataNo independent benchmark covers every model here yet
- Lowest priceSabaSaba $0.30 · Qwen2.5-VL 72B Instruct $0.93 · Aion-RP 1.0 (8B) $1.00 per 1M tokens (3:1 blend)
- Longest contextQwen2.5-VL 72B InstructQwen2.5-VL 72B Instruct 131,072 · Aion-RP 1.0 (8B) 32,768 · Saba 32,768 tokens
- Widest inputsQwen2.5-VL 72B Instruct and SabaAion-RP 1.0 (8B): Text · Qwen2.5-VL 72B Instruct: Text, Images · Saba: Text, PDFs
- Self-hostingAion-RP 1.0 (8B) and Qwen2.5-VL 72B InstructPublishes downloadable weights
| Measure | Weight | Aion-RP 1.0 (8B) | Qwen2.5-VL 72B Instruct | Saba |
|---|---|---|---|---|
| Price | 50% | 48 | 49 | 72 |
| Inputs & features | 30% | 0 | 50 | 35 |
| Context window | 20% | 0 | 25 | 0 |
| Overall | 100% | 24/100 | 44/100 | 46/100 |
Left out because at least one model lacks the data: capability. The remaining weights were rescaled.
Every spec in one table
Highlighted cells lead their row. Dashes mean the data is not published.
| Specification | Aion-RP 1.0 (8B)Aion Labs | Qwen2.5-VL 72B InstructAlibaba (Qwen) | SabaMistral AI |
|---|---|---|---|
| Capability | |||
| Capabilities Index (ECI) | — | — | — |
| ECI rank | — | — | — |
| Price per million tokens | |||
| Input | $0.80 | $0.905 | $0.20 (best) |
| Output | $1.60 | $1.00 | $0.60 (best) |
| Cached input | — | — | — |
| Blended (3:1) | $1.00 | $0.93 | $0.30 (best) |
| Long-context rate | Same rate | Same rate | Same rate |
| Price source | Median of 4 providers | Median of 6 providers | Median of 4 providers |
| Limits | |||
| Context window | 32,768 tokens | 131,072 tokens (best) | 32,768 tokens |
| Max output | 29,491 tokens (best) | 8,192 tokens | 26,214 tokens |
| Inputs and features | |||
| Text | Yes | Yes | Yes |
| Images | No | Yes | No |
| PDFs | No | No | Yes |
| Audio | No | No | No |
| Video | No | No | No |
| Reasoning | No | No | No |
| Tool calling | No | Yes | Yes |
| Structured output | No | No | No |
| Availability | |||
| Weights | Open | Open | Proprietary |
| API model ID | — | — | — |
| API providers | 4 | 6 (best) | 4 |
| Released | Feb 4, 2025 | Jan 26, 2025 | Feb 17, 2025 |
| Knowledge cutoff | Dec 31, 2023 | Apr 2024 | Sep 30, 2024 |
What would a month cost?
Enter your expected volume in millions of tokens. List prices only; caching and batch discounts would lower these.
- Aion-RP 1.0 (8B)$11.20
- Qwen2.5-VL 72B Instruct$11.06
- Saba$3.20
Which should you choose?
Which is better: Aion-RP 1.0 (8B), Qwen2.5-VL 72B Instruct or Saba?
It is close. Our weighted score puts them within 2 points (Saba 46/100, Qwen2.5-VL 72B Instruct 44/100, Aion-RP 1.0 (8B) 24/100), so choose by what matters most for your work: Saba on price and Qwen2.5-VL 72B Instruct for long inputs. The score weighs price 50%, inputs & features 30%, context window 20%. None of the shared benchmarks cover every model here yet, so capability is left out of this verdict.
Which is cheaper, Aion-RP 1.0 (8B), Qwen2.5-VL 72B Instruct or Saba?
Saba is cheaper at $0.20 input / $0.60 output per million tokens (median across 4 API providers). Qwen2.5-VL 72B Instruct costs $0.905 input / $1.00 output per million tokens (median across 6 API providers); Aion-RP 1.0 (8B) costs $0.80 input / $1.60 output per million tokens (median across 4 API providers). At a typical mix of three input tokens to one output token, that is $0.30 per million tokens for Saba versus $0.93 for Qwen2.5-VL 72B Instruct (3.1× as much) and $1.00 for Aion-RP 1.0 (8B) (3.3× as much).
Which scores higher on benchmarks?
There is no independent benchmark that covers all three models yet. Aion-RP 1.0 (8B) has not been scored yet, Qwen2.5-VL 72B Instruct has not been scored yet and Saba has not been scored yet.
Which is better for coding?
There are no published SWE-bench Verified results for Aion-RP 1.0 (8B), Qwen2.5-VL 72B Instruct and Saba yet, so there is no like-for-like coding score. Test both on a sample of your own tasks. Note that Aion-RP 1.0 (8B) does not support tool calling, which most coding agents need.
Which has the bigger context window?
Qwen2.5-VL 72B Instruct has the largest context window at 131,072 tokens, against 32,768 for Aion-RP 1.0 (8B) and 32,768 for Saba. Maximum output per response: Aion-RP 1.0 (8B) up to 29,491, Qwen2.5-VL 72B Instruct up to 8,192, Saba up to 26,214 tokens.
Which can read images, PDFs, audio or video?
Aion-RP 1.0 (8B) accepts text; Qwen2.5-VL 72B Instruct accepts text and images; Saba accepts text and PDFs. Qwen2.5-VL 72B Instruct handles the widest range of inputs.
Are any of these open source?
Aion-RP 1.0 (8B) and Qwen2.5-VL 72B Instruct publishes its weights and can be self-hosted; Saba is proprietary.
Which is newer?
Saba is the newest, released Feb 17, 2025. Aion-RP 1.0 (8B) came out Feb 4, 2025; Qwen2.5-VL 72B Instruct came out Jan 26, 2025. Knowledge cutoff: Aion-RP 1.0 (8B) Dec 31, 2023, Qwen2.5-VL 72B Instruct Apr 2024, Saba Sep 30, 2024.
How do you decide the winner?
Each model gets a 0–100 score on capability (50%, independent benchmark results); price (25%, blended price per million tokens (3 input : 1 output), log scale); inputs & features (15%, image, PDF, audio and video input, tool calling, structured output and reasoning); context window (10%, maximum tokens per request, log scale). Dimensions missing for any model are dropped and the remaining weights rescaled, so every model is judged on the same evidence. Specs and prices come from public model listings and the labs’ own API pages; capability scores come from independent benchmark runs. Data updated Nov 2, 2026.
