| Model | Context | Cost (in/out per 1M) | Strengths | Weaknesses | |-------|---------|---------------------|-----------|-------------| | | 1M | $0.80 / $2.50 | Long context, bilingual, speed | Math, formal reasoning | | DeepSeek-V3 | 128k | $0.27 / $1.10 | Cheapest, strong coding | Shorter context, no RLHF | | GPT-4o-mini | 128k | $0.15 / $0.60 | Generalist, tool use | Lower bilingual quality | | Claude 3 Haiku | 200k | $0.25 / $1.25 | Safety, low hallucination | Poor Chinese, slower | | Gemini 1.5 Flash | 1M | $0.35 / $1.05 | Very fast, native multimodal | Less accurate on complex reasoning |
As the snapped its final images of the object, scientists wondered if 3I/ATLAS was merely a rock or, as some theorized, an alien craft drifting from a distant system [28]. On Earth, the traders using Interstellar v3 didn't care about the composition of the comet. To them, the "Sirius" update was their own interstellar vessel, a powerful frontend that gave them the "most powerful and untamed" access to the market they had ever known [5].
Interstellar-v3 supports a wide range of propulsion systems, including:
: Designed to equalize airflow performance whether the fan is running in forward or reverse mode. Technical Specifications
| Benchmark | GPT-4 Turbo | Claude 3.5 Sonnet | | | :--- | :--- | :--- | :--- | | MMLU (5-shot) | 86.4% | 88.7% | 91.2% | | GSM8K (Math) | 92.0% | 95.4% | 98.6% | | HumanEval (Coding) | 84.2% | 92.0% | 96.5% | | Long Context (1M tokens) | 65% accuracy | 78% accuracy | 94% accuracy | | Vibe-Eval (Video) | N/A | 32% | 87% |
| Model | Context | Cost (in/out per 1M) | Strengths | Weaknesses | |-------|---------|---------------------|-----------|-------------| | | 1M | $0.80 / $2.50 | Long context, bilingual, speed | Math, formal reasoning | | DeepSeek-V3 | 128k | $0.27 / $1.10 | Cheapest, strong coding | Shorter context, no RLHF | | GPT-4o-mini | 128k | $0.15 / $0.60 | Generalist, tool use | Lower bilingual quality | | Claude 3 Haiku | 200k | $0.25 / $1.25 | Safety, low hallucination | Poor Chinese, slower | | Gemini 1.5 Flash | 1M | $0.35 / $1.05 | Very fast, native multimodal | Less accurate on complex reasoning |
As the snapped its final images of the object, scientists wondered if 3I/ATLAS was merely a rock or, as some theorized, an alien craft drifting from a distant system [28]. On Earth, the traders using Interstellar v3 didn't care about the composition of the comet. To them, the "Sirius" update was their own interstellar vessel, a powerful frontend that gave them the "most powerful and untamed" access to the market they had ever known [5].
Interstellar-v3 supports a wide range of propulsion systems, including:
: Designed to equalize airflow performance whether the fan is running in forward or reverse mode. Technical Specifications
| Benchmark | GPT-4 Turbo | Claude 3.5 Sonnet | | | :--- | :--- | :--- | :--- | | MMLU (5-shot) | 86.4% | 88.7% | 91.2% | | GSM8K (Math) | 92.0% | 95.4% | 98.6% | | HumanEval (Coding) | 84.2% | 92.0% | 96.5% | | Long Context (1M tokens) | 65% accuracy | 78% accuracy | 94% accuracy | | Vibe-Eval (Video) | N/A | 32% | 87% |