Model Comparison
Kimi K3 vs GPT-5.6 Terra
Kimi K3 (max) leads GPT-5.6 Terra (ultra) by 129 Elo, and their 95% ranges sit 45 Elo apart. Among 78 models, each ranked by its best configuration, Kimi K3 is #10 and GPT-5.6 Terra is #13. Cost per script: $0.362 for Kimi K3 (max) and $0.275 for GPT-5.6 Terra (ultra).
Last updated , from board data generated at 01:13 UTC.
Picks and Verdict
PickKimi K3
Kimi K3 (max)
Pick Kimi K3 (max) as the one setting here clear of the pack, 129 Elo above Terra's best for $0.362 a script.
PickGPT-5.6 Terra
GPT-5.6 Terra (medium)
Pick GPT-5.6 Terra (medium) for cheap drafts at $0.062 a script, since Kimi K3 (high) and Terra's own ultra setting cost more and the board cannot separate them from it.
Only one setting on this page pulls away from the others, and it's Kimi K3 (max). It sits 129 Elo above GPT-5.6 Terra (ultra), with 95% ranges that don't overlap: the #10 model of 78 against the #13 model, each ranked by its best setting. Below that, Kimi K3 (high) and Terra at medium land 20 Elo apart with overlapping ranges. Terra at ultra sits clear of Kimi at high, yet only 76 Elo above its own medium setting, with ranges that overlap. The bill differs: $0.128 for Kimi at high, $0.275 for Terra at ultra and $0.062 at medium. Terra's ultra setting costs more than four times as much as medium for a gain the board can't confirm, so I wouldn't pay for it. Medium even sounds closer to me, by almost two points on tone. Ultra does keep the best substance score here, almost three points above Kimi at max. Kimi's own step from high to max goes the other way, 225 Elo for nearly three times the price. I'd run Terra at medium for drafts and pay $0.362 for Kimi at max on the final version.
Disclosure: GPT-5.6 Terra comes from the same lab as one of the four judges, GPT-6.1 Sol. The other three sit outside its lab, so it is never scored by its own lab alone (how the panel works). The judge analysis measures how much each judge favors its own lab.
How to Read the Writing Score, Writing Elo and 95% Ranges
Writing Score and Writing Elo. The Writing Score (0 to 100) is the weighted average of the nine metric scores: seven from the judges, Length Adherence from code and the Slop Score half from each. The Writing Elo is a rating from pairwise comparisons: on each script, each configuration's average across its drafts is compared with every other configuration's, and gaps within the draft-to-draft variation count as draws. Elo rewards winning often, so the two can put configurations in a different order.
Elo ranges. Each 95% range shows how far a configuration's Elo could move if the drafts had come out slightly differently. It comes from resampling every configuration's recorded draft scores on each reference script and ranking again. Comparing two separate ranges does not test the Elo gap between them: ranges far apart are strong evidence for the order, and overlapping ranges do not mean a tie. The reference scripts stay fixed, so the ranges say nothing about new scripts. The methodology page has the full description.
All Thinking Levels of Both Models
Every ranked configuration of Kimi K3 and GPT-5.6 Terra, scored on the same 10 scripts by the same four judges.
- Each row is one configuration: the model at one thinking level. The dot is its Writing Elo, a rating from pairwise comparisons of the scores on each script, and the bar is its 95% range: how far that Elo could move if the drafts had come out slightly differently.
- When two bars overlap, the board cannot order those two configurations. Bars far apart are strong evidence for the order.
- The table under the chart lists the same configurations in the same order and adds the Writing Score (the 0 to 100 consensus of the judges and code), cost, time and configuration rank (the place among all configurations on the board).
Shaded rows are the configurations chosen for the metric comparison below.
| Configuration | Writing Elo | 95% Range | Writing Score | Cost / Script | Time / Script | Reasoning Tokens | Configuration Rank |
|---|---|---|---|---|---|---|---|
| Kimi K3 (max) | 2203 | 2158 to 2243 | 82.2 | $0.362 | 9.9 min | 18,568 | #20 |
| GPT-5.6 Terra (ultra) | 2074 | 2035 to 2113 | 81.4 | $0.275 | 8.0 min | N/A | #25 |
| GPT-5.6 Terra (medium) | 1998 | 1952 to 2041 | 80.7 | $0.062 | 1.4 min | N/A | #33 |
| Kimi K3 (high) | 1978 | 1936 to 2029 | 80.3 | $0.128 | 2.8 min | 3,081 | #37 |
Cost is one script at the API list price of the route used, uncached and without judging. Time is the average for one accepted script, retries included. N/A under Reasoning Tokens means the board has no separate reasoning count for that route.
Computed Findings Across Thinking Levels
Each finding is computed from the configurations above. Where two 95% ranges overlap, the board cannot order those two configurations.
- Best Configurations Head to Head. Kimi K3 (max) leads GPT-5.6 Terra (ultra) by 129 Elo, and their 95% ranges do not overlap (2158 to 2243 against 2035 to 2113), which is strong evidence for that order.
- Lowest Thinking Levels Head to Head. At their lowest thinking levels, GPT-5.6 Terra (medium) at Elo 1998 and $0.062 per script leads Kimi K3 (high) at Elo 1978 and $0.128 per script by 20 Elo. Their 95% ranges overlap, so the board cannot separate them.
- Cheapest Kimi K3 Configuration Above GPT-5.6 Terra's Best. The cheapest Kimi K3 configuration with a higher Writing Elo than GPT-5.6 Terra (ultra), at Elo 2074 and $0.275 per script, is Kimi K3 (max), at Elo 2203 and $0.362 per script. It costs about 1.3 times as much per script.
- Lowest to Highest Thinking Level. Kimi K3 moves +225 Elo from high to max, at about 2.8 times the cost per script. GPT-5.6 Terra moves +76 Elo from medium to ultra, at about 4.4 times the cost per script.
Two Chosen Configurations Compared
Pick one configuration of each model to compare their metric scores, consistency across drafts and per-script scores. Each menu starts on the model's best configuration.
Scores by Metric
Kimi K3 (max) scores higher on six of the nine metrics and GPT-5.6 Terra (ultra) on three. Weighted by each metric's share of the Writing Score, Tone & Voice Match does the most to put Kimi K3 (max) ahead, with a 1.8-point lead on that metric. The weight beside each metric is its share of the Writing Score.
Consistency Across Drafts
| Kimi K3 (max) | GPT-5.6 Terra (ultra) | |
|---|---|---|
| Score spread across drafts (±) | 1.68 | 1.38 |
| Lowest script score | 80.0 Script 10 | 79.8 Script 4 |
| Highest script score | 83.8 Script 9 | 83.0 Script 6 |
| Range across scripts | 3.8 | 3.2 |
Spread is how much the Writing Score moves between a configuration's drafts, and lower is steadier. Green marks the better number in each row.
Scores by Script
Writing Score on each reference script. Gap is Kimi K3 (max) minus GPT-5.6 Terra (ultra), and the higher score is in bold.
| Script | Kimi K3 (max) | GPT-5.6 Terra (ultra) | Gap |
|---|---|---|---|
| Script 1 opinion / warning explainer | 81.4 | 81.1 | +0.3 |
| Script 2 news-analysis / skeptical explainer | 82.6 | 82.1 | +0.5 |
| Script 3 personal roadmap / opinion | 81.8 | 82.5 | -0.7 |
| Script 4 short explainer | 80.1 | 79.8 | +0.3 |
| Script 5 news-analysis / opinion explainer | 83.4 | 82.2 | +1.2 |
| Script 6 founder announcement / personal origin story | 83.2 | 83.0 | +0.2 |
| Script 7 personal technical walkthrough / agentic workflow case study | 83.6 | 80.6 | +3.0 |
| Script 8 product release announcement / personal observation | 82.3 | 80.3 | +2.0 |
| Script 9 career guide / hiring analysis | 83.8 | 81.8 | +2.0 |
| Script 10 engineering process walkthrough / presentation adaptation | 80.0 | 80.6 | -0.6 |
Scorecards and Thinking Pages
Each configuration name in the table above opens its own scorecard.
Compare With Another Model
Pick a second model. A pair with its own written match-up opens that page, and any other pair opens in the comparison tool.
Nearby Match-Ups by Best Writing Elo
Every configuration wrote five drafts for each of the same 10 briefs, from the same prompt. Generation was closed-book, with the prompt as the only source and no web search or other tools. Each judge scores every draft once: Claude Opus 5.5 at medium effort through the Claude Code subscription, GPT-6.1 Sol at medium reasoning through the Codex CLI, GLM 5.3 Flash through OpenRouter, and Jev 1.13, a typed judge from TypeSafe that answers each metric as a multiple-choice question. The channel's finished script is their reference for the voice, and it gets no score on the board itself.
← All comparisons